Renaming of servers for coherency.

VisualServer -> RenderingServer
PhysicsServer -> PhysicsServer3D
Physics2DServer -> PhysicsServer2D
NavigationServer -> NavigationServer3D
Navigation2DServer -> NavigationServer2D

Also renamed corresponding files.

servers/rendering/rasterizer_rd/SCsub

@@ -0,0 +1,7 @@
+#!/usr/bin/env python
+
+Import('env')
+
+env.add_source_files(env.servers_sources, "*.cpp")
+
+SConscript("shaders/SCsub")

servers/rendering/rasterizer_rd/light_cluster_builder.cpp

@@ -0,0 +1,255 @@
+/*************************************************************************/
+/*  light_cluster_builder.cpp                                            */
+/*************************************************************************/
+/*                       This file is part of:                           */
+/*                           GODOT ENGINE                                */
+/*                      https://godotengine.org                          */
+/*************************************************************************/
+/* Copyright (c) 2007-2020 Juan Linietsky, Ariel Manzur.                 */
+/* Copyright (c) 2014-2020 Godot Engine contributors (cf. AUTHORS.md).   */
+/*                                                                       */
+/* Permission is hereby granted, free of charge, to any person obtaining */
+/* a copy of this software and associated documentation files (the       */
+/* "Software"), to deal in the Software without restriction, including   */
+/* without limitation the rights to use, copy, modify, merge, publish,   */
+/* distribute, sublicense, and/or sell copies of the Software, and to    */
+/* permit persons to whom the Software is furnished to do so, subject to */
+/* the following conditions:                                             */
+/*                                                                       */
+/* The above copyright notice and this permission notice shall be        */
+/* included in all copies or substantial portions of the Software.       */
+/*                                                                       */
+/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,       */
+/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF    */
+/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
+/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY  */
+/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,  */
+/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE     */
+/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                */
+/*************************************************************************/
+
+#include "light_cluster_builder.h"
+
+void LightClusterBuilder::begin(const Transform &p_view_transform, const CameraMatrix &p_cam_projection) {
+	view_xform = p_view_transform;
+	projection = p_cam_projection;
+	z_near = -projection.get_z_near();
+	z_far = -projection.get_z_far();
+
+	//reset counts
+	light_count = 0;
+	refprobe_count = 0;
+	item_count = 0;
+	sort_id_count = 0;
+}
+
+void LightClusterBuilder::bake_cluster() {
+
+	float slice_depth = (z_near - z_far) / depth;
+
+	uint8_t *cluster_dataw = cluster_data.ptrw();
+	Cell *cluster_data_ptr = (Cell *)cluster_dataw;
+	//clear the cluster
+	zeromem(cluster_data_ptr, (width * height * depth * sizeof(Cell)));
+
+	/* Step 1, create cell positions and count them */
+
+	for (uint32_t i = 0; i < item_count; i++) {
+
+		const Item &item = items[i];
+
+		int from_slice = Math::floor((z_near - (item.aabb.position.z + item.aabb.size.z)) / slice_depth);
+		int to_slice = Math::floor((z_near - item.aabb.position.z) / slice_depth);
+
+		if (from_slice >= (int)depth || to_slice < 0) {
+			continue; //sorry no go
+		}
+
+		from_slice = MAX(0, from_slice);
+		to_slice = MIN((int)depth - 1, to_slice);
+
+		for (int j = from_slice; j <= to_slice; j++) {
+
+			Vector3 min = item.aabb.position;
+			Vector3 max = item.aabb.position + item.aabb.size;
+
+			float limit_near = MIN((z_near - slice_depth * j), max.z);
+			float limit_far = MAX((z_near - slice_depth * (j + 1)), min.z);
+
+			max.z = limit_near;
+			min.z = limit_near;
+
+			Vector3 proj_min = projection.xform(min);
+			Vector3 proj_max = projection.xform(max);
+
+			int near_from_x = int(Math::floor((proj_min.x * 0.5 + 0.5) * width));
+			int near_from_y = int(Math::floor((-proj_max.y * 0.5 + 0.5) * height));
+			int near_to_x = int(Math::floor((proj_max.x * 0.5 + 0.5) * width));
+			int near_to_y = int(Math::floor((-proj_min.y * 0.5 + 0.5) * height));
+
+			max.z = limit_far;
+			min.z = limit_far;
+
+			proj_min = projection.xform(min);
+			proj_max = projection.xform(max);
+
+			int far_from_x = int(Math::floor((proj_min.x * 0.5 + 0.5) * width));
+			int far_from_y = int(Math::floor((-proj_max.y * 0.5 + 0.5) * height));
+			int far_to_x = int(Math::floor((proj_max.x * 0.5 + 0.5) * width));
+			int far_to_y = int(Math::floor((-proj_min.y * 0.5 + 0.5) * height));
+
+			//print_line(itos(j) + " near - " + Vector2i(near_from_x, near_from_y) + " -> " + Vector2i(near_to_x, near_to_y));
+			//print_line(itos(j) + " far - " + Vector2i(far_from_x, far_from_y) + " -> " + Vector2i(far_to_x, far_to_y));
+
+			int from_x = MIN(near_from_x, far_from_x);
+			int from_y = MIN(near_from_y, far_from_y);
+			int to_x = MAX(near_to_x, far_to_x);
+			int to_y = MAX(near_to_y, far_to_y);
+
+			if (from_x >= (int)width || to_x < 0 || from_y >= (int)height || to_y < 0) {
+				continue;
+			}
+
+			int sx = MAX(0, from_x);
+			int sy = MAX(0, from_y);
+			int dx = MIN((int)width - 1, to_x);
+			int dy = MIN((int)height - 1, to_y);
+
+			//print_line(itos(j) + " - " + Vector2i(sx, sy) + " -> " + Vector2i(dx, dy));
+
+			for (int x = sx; x <= dx; x++) {
+				for (int y = sy; y <= dy; y++) {
+					uint32_t offset = j * (width * height) + y * width + x;
+
+					if (unlikely(sort_id_count == sort_id_max)) {
+						sort_id_max = nearest_power_of_2_templated(sort_id_max + 1);
+						sort_ids = (SortID *)memrealloc(sort_ids, sizeof(SortID) * sort_id_max);
+						if (ids.size()) {
+
+							ids.resize(sort_id_max);
+							RD::get_singleton()->free(items_buffer);
+							items_buffer = RD::get_singleton()->storage_buffer_create(sizeof(uint32_t) * sort_id_max);
+						}
+					}
+
+					sort_ids[sort_id_count].cell_index = offset;
+					sort_ids[sort_id_count].item_index = item.index;
+					sort_ids[sort_id_count].item_type = item.type;
+
+					sort_id_count++;
+
+					//for now, only count
+					cluster_data_ptr[offset].item_pointers[item.type]++;
+					//print_line("at offset " + itos(offset) + " value: " + itos(cluster_data_ptr[offset].item_pointers[item.type]));
+				}
+			}
+		}
+	}
+
+	/* Step 2, Assign pointers (and reset counters) */
+
+	uint32_t offset = 0;
+	for (uint32_t i = 0; i < (width * height * depth); i++) {
+		for (int j = 0; j < ITEM_TYPE_MAX; j++) {
+			uint32_t count = cluster_data_ptr[i].item_pointers[j]; //save count
+			cluster_data_ptr[i].item_pointers[j] = offset; //replace count by pointer
+			offset += count; //increase offset by count;
+		}
+	}
+
+	//print_line("offset: " + itos(offset));
+	/* Step 3, Place item lists */
+
+	uint32_t *ids_ptr = ids.ptrw();
+
+	for (uint32_t i = 0; i < sort_id_count; i++) {
+		const SortID &id = sort_ids[i];
+		Cell &cell = cluster_data_ptr[id.cell_index];
+		uint32_t pointer = cell.item_pointers[id.item_type] & POINTER_MASK;
+		uint32_t counter = cell.item_pointers[id.item_type] >> COUNTER_SHIFT;
+		ids_ptr[pointer + counter] = id.item_index;
+
+		cell.item_pointers[id.item_type] = pointer | ((counter + 1) << COUNTER_SHIFT);
+	}
+
+	RD::get_singleton()->texture_update(cluster_texture, 0, cluster_data, true);
+	RD::get_singleton()->buffer_update(items_buffer, 0, offset * sizeof(uint32_t), ids_ptr, true);
+}
+
+void LightClusterBuilder::setup(uint32_t p_width, uint32_t p_height, uint32_t p_depth) {
+
+	if (width == p_width && height == p_height && depth == p_depth) {
+		return;
+	}
+	if (cluster_texture.is_valid()) {
+		RD::get_singleton()->free(cluster_texture);
+	}
+
+	width = p_width;
+	height = p_height;
+	depth = p_depth;
+
+	cluster_data.resize(width * height * depth * sizeof(Cell));
+
+	{
+		RD::TextureFormat tf;
+		tf.format = RD::DATA_FORMAT_R32G32B32A32_UINT;
+		tf.type = RD::TEXTURE_TYPE_3D;
+		tf.width = width;
+		tf.height = height;
+		tf.depth = depth;
+		tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_UPDATE_BIT;
+
+		cluster_texture = RD::get_singleton()->texture_create(tf, RD::TextureView());
+	}
+}
+
+RID LightClusterBuilder::get_cluster_texture() const {
+	return cluster_texture;
+}
+RID LightClusterBuilder::get_cluster_indices_buffer() const {
+	return items_buffer;
+}
+
+LightClusterBuilder::LightClusterBuilder() {
+	//initialize accumulators to something
+	lights = (LightData *)memalloc(sizeof(LightData) * 1024);
+	light_max = 1024;
+
+	refprobes = (OrientedBoxData *)memalloc(sizeof(OrientedBoxData) * 1024);
+	refprobe_max = 1024;
+
+	decals = (OrientedBoxData *)memalloc(sizeof(OrientedBoxData) * 1024);
+	decal_max = 1024;
+
+	items = (Item *)memalloc(sizeof(Item) * 1024);
+	item_max = 1024;
+
+	sort_ids = (SortID *)memalloc(sizeof(SortID) * 1024);
+	ids.resize(2014);
+	items_buffer = RD::get_singleton()->storage_buffer_create(sizeof(uint32_t) * 1024);
+	item_max = 1024;
+}
+LightClusterBuilder::~LightClusterBuilder() {
+
+	if (cluster_data.size()) {
+		RD::get_singleton()->free(cluster_texture);
+	}
+
+	if (lights) {
+		memfree(lights);
+	}
+	if (refprobes) {
+		memfree(refprobes);
+	}
+	if (decals) {
+		memfree(decals);
+	}
+	if (items) {
+		memfree(items);
+	}
+	if (sort_ids) {
+		memfree(sort_ids);
+		RD::get_singleton()->free(items_buffer);
+	}
+}

servers/rendering/rasterizer_rd/light_cluster_builder.h

@@ -0,0 +1,291 @@
+/*************************************************************************/
+/*  light_cluster_builder.h                                              */
+/*************************************************************************/
+/*                       This file is part of:                           */
+/*                           GODOT ENGINE                                */
+/*                      https://godotengine.org                          */
+/*************************************************************************/
+/* Copyright (c) 2007-2020 Juan Linietsky, Ariel Manzur.                 */
+/* Copyright (c) 2014-2020 Godot Engine contributors (cf. AUTHORS.md).   */
+/*                                                                       */
+/* Permission is hereby granted, free of charge, to any person obtaining */
+/* a copy of this software and associated documentation files (the       */
+/* "Software"), to deal in the Software without restriction, including   */
+/* without limitation the rights to use, copy, modify, merge, publish,   */
+/* distribute, sublicense, and/or sell copies of the Software, and to    */
+/* permit persons to whom the Software is furnished to do so, subject to */
+/* the following conditions:                                             */
+/*                                                                       */
+/* The above copyright notice and this permission notice shall be        */
+/* included in all copies or substantial portions of the Software.       */
+/*                                                                       */
+/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,       */
+/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF    */
+/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
+/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY  */
+/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,  */
+/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE     */
+/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                */
+/*************************************************************************/
+
+#ifndef LIGHT_CLUSTER_BUILDER_H
+#define LIGHT_CLUSTER_BUILDER_H
+
+#include "servers/rendering/rasterizer_rd/rasterizer_storage_rd.h"
+
+class LightClusterBuilder {
+public:
+	enum LightType {
+		LIGHT_TYPE_OMNI,
+		LIGHT_TYPE_SPOT
+	};
+
+	enum ItemType {
+		ITEM_TYPE_OMNI_LIGHT,
+		ITEM_TYPE_SPOT_LIGHT,
+		ITEM_TYPE_REFLECTION_PROBE,
+		ITEM_TYPE_DECAL,
+		ITEM_TYPE_MAX //should always be 4
+	};
+
+	enum {
+		COUNTER_SHIFT = 20, //one million total ids
+		POINTER_MASK = (1 << COUNTER_SHIFT) - 1,
+		COUNTER_MASK = 0xfff // 4096 items per cell
+	};
+
+private:
+	struct LightData {
+		float position[3];
+		uint32_t type;
+		float radius;
+		float spot_aperture;
+		uint32_t pad[2];
+	};
+
+	uint32_t light_count = 0;
+	uint32_t light_max = 0;
+	LightData *lights = nullptr;
+
+	struct OrientedBoxData {
+		float position[3];
+		uint32_t pad;
+		float x_axis[3];
+		uint32_t pad2;
+		float y_axis[3];
+		uint32_t pad3;
+		float z_axis[3];
+		uint32_t pad4;
+	};
+
+	uint32_t refprobe_count = 0;
+	uint32_t refprobe_max = 0;
+	OrientedBoxData *refprobes = nullptr;
+
+	uint32_t decal_count = 0;
+	uint32_t decal_max = 0;
+	OrientedBoxData *decals = nullptr;
+
+	struct Item {
+		AABB aabb;
+		ItemType type;
+		uint32_t index;
+	};
+
+	Item *items = nullptr;
+	uint32_t item_count = 0;
+	uint32_t item_max = 0;
+
+	uint32_t width = 0;
+	uint32_t height = 0;
+	uint32_t depth = 0;
+
+	struct Cell {
+		uint32_t item_pointers[ITEM_TYPE_MAX];
+	};
+
+	Vector<uint8_t> cluster_data;
+	RID cluster_texture;
+
+	struct SortID {
+		uint32_t cell_index;
+		uint32_t item_index;
+		ItemType item_type;
+	};
+
+	SortID *sort_ids = nullptr;
+	Vector<uint32_t> ids;
+	uint32_t sort_id_count = 0;
+	uint32_t sort_id_max = 0;
+	RID items_buffer;
+
+	Transform view_xform;
+	CameraMatrix projection;
+	float z_far = 0;
+	float z_near = 0;
+
+	_FORCE_INLINE_ void _add_item(const AABB &p_aabb, ItemType p_type, uint32_t p_index) {
+		if (unlikely(item_count == item_max)) {
+			item_max = nearest_power_of_2_templated(item_max + 1);
+			items = (Item *)memrealloc(items, sizeof(Item) * item_max);
+		}
+
+		Item &item = items[item_count];
+		item.aabb = p_aabb;
+		item.index = p_index;
+		item.type = p_type;
+		item_count++;
+	}
+
+public:
+	void begin(const Transform &p_view_transform, const CameraMatrix &p_cam_projection);
+
+	_FORCE_INLINE_ void add_light(LightType p_type, const Transform &p_transform, float p_radius, float p_spot_aperture) {
+		if (unlikely(light_count == light_max)) {
+			light_max = nearest_power_of_2_templated(light_max + 1);
+			lights = (LightData *)memrealloc(lights, sizeof(LightData) * light_max);
+		}
+
+		LightData &ld = lights[light_count];
+		ld.type = p_type;
+		ld.position[0] = p_transform.origin.x;
+		ld.position[1] = p_transform.origin.y;
+		ld.position[2] = p_transform.origin.z;
+		ld.radius = p_radius;
+		ld.spot_aperture = p_spot_aperture;
+
+		Transform xform = view_xform * p_transform;
+
+		ld.radius *= xform.basis.get_uniform_scale();
+
+		AABB aabb;
+
+		switch (p_type) {
+			case LIGHT_TYPE_OMNI: {
+				aabb.position = xform.origin;
+				aabb.size = Vector3(ld.radius, ld.radius, ld.radius);
+				aabb.position -= aabb.size;
+				aabb.size *= 2.0;
+
+				_add_item(aabb, ITEM_TYPE_OMNI_LIGHT, light_count);
+			} break;
+			case LIGHT_TYPE_SPOT: {
+				Vector3 v(0, 0, -1);
+				v.rotated(Vector3(0, 1, 0), Math::deg2rad(ld.spot_aperture)); //rotate in x-z
+				v.normalize();
+				v *= ld.radius;
+				v.y = v.x;
+
+				aabb.position = xform.origin;
+				aabb.expand_to(xform.xform(v));
+				aabb.expand_to(xform.xform(Vector3(-v.x, v.y, v.z)));
+				aabb.expand_to(xform.xform(Vector3(-v.x, -v.y, v.z)));
+				aabb.expand_to(xform.xform(Vector3(v.x, -v.y, v.z)));
+				_add_item(aabb, ITEM_TYPE_SPOT_LIGHT, light_count);
+			} break;
+		}
+
+		light_count++;
+	}
+
+	_FORCE_INLINE_ void add_reflection_probe(const Transform &p_transform, const Vector3 &p_half_extents) {
+
+		if (unlikely(refprobe_count == refprobe_max)) {
+			refprobe_max = nearest_power_of_2_templated(refprobe_max + 1);
+			refprobes = (OrientedBoxData *)memrealloc(refprobes, sizeof(OrientedBoxData) * refprobe_max);
+		}
+
+		OrientedBoxData &rp = refprobes[refprobe_count];
+		Vector3 origin = p_transform.origin;
+		rp.position[0] = origin.x;
+		rp.position[1] = origin.y;
+		rp.position[2] = origin.z;
+
+		Vector3 x_axis = p_transform.basis.get_axis(0) * p_half_extents.x;
+		rp.x_axis[0] = x_axis.x;
+		rp.x_axis[1] = x_axis.y;
+		rp.x_axis[2] = x_axis.z;
+
+		Vector3 y_axis = p_transform.basis.get_axis(1) * p_half_extents.y;
+		rp.y_axis[0] = y_axis.x;
+		rp.y_axis[1] = y_axis.y;
+		rp.y_axis[2] = y_axis.z;
+
+		Vector3 z_axis = p_transform.basis.get_axis(2) * p_half_extents.z;
+		rp.z_axis[0] = z_axis.x;
+		rp.z_axis[1] = z_axis.y;
+		rp.z_axis[2] = z_axis.z;
+
+		AABB aabb;
+
+		aabb.position = origin + x_axis + y_axis + z_axis;
+		aabb.expand_to(origin + x_axis + y_axis - z_axis);
+		aabb.expand_to(origin + x_axis - y_axis + z_axis);
+		aabb.expand_to(origin + x_axis - y_axis - z_axis);
+		aabb.expand_to(origin - x_axis + y_axis + z_axis);
+		aabb.expand_to(origin - x_axis + y_axis - z_axis);
+		aabb.expand_to(origin - x_axis - y_axis + z_axis);
+		aabb.expand_to(origin - x_axis - y_axis - z_axis);
+
+		_add_item(aabb, ITEM_TYPE_REFLECTION_PROBE, refprobe_count);
+
+		refprobe_count++;
+	}
+
+	_FORCE_INLINE_ void add_decal(const Transform &p_transform, const Vector2 &p_half_extents, float p_depth) {
+
+		if (unlikely(decal_count == decal_max)) {
+			decal_max = nearest_power_of_2_templated(decal_max + 1);
+			decals = (OrientedBoxData *)memrealloc(decals, sizeof(OrientedBoxData) * decal_max);
+		}
+
+		OrientedBoxData &dc = decals[decal_count];
+
+		Vector3 z_axis = -p_transform.basis.get_axis(2) * p_depth * 0.5;
+		dc.z_axis[0] = z_axis.x;
+		dc.z_axis[1] = z_axis.y;
+		dc.z_axis[2] = z_axis.z;
+
+		Vector3 origin = p_transform.origin - z_axis;
+		dc.position[0] = origin.x;
+		dc.position[1] = origin.y;
+		dc.position[2] = origin.z;
+
+		Vector3 x_axis = p_transform.basis.get_axis(0) * p_half_extents.x;
+		dc.x_axis[0] = x_axis.x;
+		dc.x_axis[1] = x_axis.y;
+		dc.x_axis[2] = x_axis.z;
+
+		Vector3 y_axis = p_transform.basis.get_axis(1) * p_half_extents.y;
+		dc.y_axis[0] = y_axis.x;
+		dc.y_axis[1] = y_axis.y;
+		dc.y_axis[2] = y_axis.z;
+
+		AABB aabb;
+
+		aabb.position = origin + x_axis + y_axis + z_axis;
+		aabb.expand_to(origin + x_axis + y_axis - z_axis);
+		aabb.expand_to(origin + x_axis - y_axis + z_axis);
+		aabb.expand_to(origin + x_axis - y_axis - z_axis);
+		aabb.expand_to(origin - x_axis + y_axis + z_axis);
+		aabb.expand_to(origin - x_axis + y_axis - z_axis);
+		aabb.expand_to(origin - x_axis - y_axis + z_axis);
+		aabb.expand_to(origin - x_axis - y_axis - z_axis);
+
+		_add_item(aabb, ITEM_TYPE_DECAL, decal_count);
+
+		decal_count++;
+	}
+
+	void bake_cluster();
+
+	void setup(uint32_t p_width, uint32_t p_height, uint32_t p_depth);
+
+	RID get_cluster_texture() const;
+	RID get_cluster_indices_buffer() const;
+
+	LightClusterBuilder();
+	~LightClusterBuilder();
+};
+
+#endif // LIGHT_CLUSTER_BUILDER_H

servers/rendering/rasterizer_rd/rasterizer_canvas_rd.h

@@ -0,0 +1,501 @@
+/*************************************************************************/
+/*  rasterizer_canvas_rd.h                                               */
+/*************************************************************************/
+/*                       This file is part of:                           */
+/*                           GODOT ENGINE                                */
+/*                      https://godotengine.org                          */
+/*************************************************************************/
+/* Copyright (c) 2007-2020 Juan Linietsky, Ariel Manzur.                 */
+/* Copyright (c) 2014-2020 Godot Engine contributors (cf. AUTHORS.md).   */
+/*                                                                       */
+/* Permission is hereby granted, free of charge, to any person obtaining */
+/* a copy of this software and associated documentation files (the       */
+/* "Software"), to deal in the Software without restriction, including   */
+/* without limitation the rights to use, copy, modify, merge, publish,   */
+/* distribute, sublicense, and/or sell copies of the Software, and to    */
+/* permit persons to whom the Software is furnished to do so, subject to */
+/* the following conditions:                                             */
+/*                                                                       */
+/* The above copyright notice and this permission notice shall be        */
+/* included in all copies or substantial portions of the Software.       */
+/*                                                                       */
+/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,       */
+/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF    */
+/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
+/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY  */
+/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,  */
+/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE     */
+/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                */
+/*************************************************************************/
+
+#ifndef RASTERIZER_CANVAS_RD_H
+#define RASTERIZER_CANVAS_RD_H
+
+#include "servers/rendering/rasterizer.h"
+#include "servers/rendering/rasterizer_rd/rasterizer_storage_rd.h"
+#include "servers/rendering/rasterizer_rd/render_pipeline_vertex_format_cache_rd.h"
+#include "servers/rendering/rasterizer_rd/shader_compiler_rd.h"
+#include "servers/rendering/rasterizer_rd/shaders/canvas.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/canvas_occlusion.glsl.gen.h"
+#include "servers/rendering/rendering_device.h"
+
+class RasterizerCanvasRD : public RasterizerCanvas {
+
+	RasterizerStorageRD *storage;
+
+	enum ShaderVariant {
+		SHADER_VARIANT_QUAD,
+		SHADER_VARIANT_NINEPATCH,
+		SHADER_VARIANT_PRIMITIVE,
+		SHADER_VARIANT_PRIMITIVE_POINTS,
+		SHADER_VARIANT_ATTRIBUTES,
+		SHADER_VARIANT_ATTRIBUTES_POINTS,
+		SHADER_VARIANT_QUAD_LIGHT,
+		SHADER_VARIANT_NINEPATCH_LIGHT,
+		SHADER_VARIANT_PRIMITIVE_LIGHT,
+		SHADER_VARIANT_PRIMITIVE_POINTS_LIGHT,
+		SHADER_VARIANT_ATTRIBUTES_LIGHT,
+		SHADER_VARIANT_ATTRIBUTES_POINTS_LIGHT,
+		SHADER_VARIANT_MAX
+	};
+
+	enum {
+		FLAGS_INSTANCING_STRIDE_MASK = 0xF,
+		FLAGS_INSTANCING_ENABLED = (1 << 4),
+		FLAGS_INSTANCING_HAS_COLORS = (1 << 5),
+		FLAGS_INSTANCING_COLOR_8BIT = (1 << 6),
+		FLAGS_INSTANCING_HAS_CUSTOM_DATA = (1 << 7),
+		FLAGS_INSTANCING_CUSTOM_DATA_8_BIT = (1 << 8),
+
+		FLAGS_CLIP_RECT_UV = (1 << 9),
+		FLAGS_TRANSPOSE_RECT = (1 << 10),
+		FLAGS_USING_LIGHT_MASK = (1 << 11),
+
+		FLAGS_NINEPACH_DRAW_CENTER = (1 << 12),
+		FLAGS_USING_PARTICLES = (1 << 13),
+		FLAGS_USE_PIXEL_SNAP = (1 << 14),
+
+		FLAGS_USE_SKELETON = (1 << 15),
+		FLAGS_NINEPATCH_H_MODE_SHIFT = 16,
+		FLAGS_NINEPATCH_V_MODE_SHIFT = 18,
+		FLAGS_LIGHT_COUNT_SHIFT = 20,
+
+		FLAGS_DEFAULT_NORMAL_MAP_USED = (1 << 26),
+		FLAGS_DEFAULT_SPECULAR_MAP_USED = (1 << 27)
+
+	};
+
+	enum {
+		LIGHT_FLAGS_TEXTURE_MASK = 0xFFFF,
+		LIGHT_FLAGS_BLEND_SHIFT = 16,
+		LIGHT_FLAGS_BLEND_MASK = (3 << 16),
+		LIGHT_FLAGS_BLEND_MODE_ADD = (0 << 16),
+		LIGHT_FLAGS_BLEND_MODE_SUB = (1 << 16),
+		LIGHT_FLAGS_BLEND_MODE_MIX = (2 << 16),
+		LIGHT_FLAGS_BLEND_MODE_MASK = (3 << 16),
+		LIGHT_FLAGS_HAS_SHADOW = (1 << 20),
+		LIGHT_FLAGS_FILTER_SHIFT = 22
+
+	};
+
+	enum {
+		MAX_RENDER_ITEMS = 256 * 1024,
+		MAX_LIGHT_TEXTURES = 1024,
+		DEFAULT_MAX_LIGHTS_PER_ITEM = 16,
+		DEFAULT_MAX_LIGHTS_PER_RENDER = 256
+	};
+
+	/****************/
+	/**** SHADER ****/
+	/****************/
+
+	enum PipelineVariant {
+		PIPELINE_VARIANT_QUAD,
+		PIPELINE_VARIANT_NINEPATCH,
+		PIPELINE_VARIANT_PRIMITIVE_TRIANGLES,
+		PIPELINE_VARIANT_PRIMITIVE_LINES,
+		PIPELINE_VARIANT_PRIMITIVE_POINTS,
+		PIPELINE_VARIANT_ATTRIBUTE_TRIANGLES,
+		PIPELINE_VARIANT_ATTRIBUTE_TRIANGLE_STRIP,
+		PIPELINE_VARIANT_ATTRIBUTE_LINES,
+		PIPELINE_VARIANT_ATTRIBUTE_LINES_STRIP,
+		PIPELINE_VARIANT_ATTRIBUTE_POINTS,
+		PIPELINE_VARIANT_MAX
+	};
+	enum PipelineLightMode {
+		PIPELINE_LIGHT_MODE_DISABLED,
+		PIPELINE_LIGHT_MODE_ENABLED,
+		PIPELINE_LIGHT_MODE_MAX
+	};
+
+	struct PipelineVariants {
+		RenderPipelineVertexFormatCacheRD variants[PIPELINE_LIGHT_MODE_MAX][PIPELINE_VARIANT_MAX];
+	};
+
+	struct {
+		CanvasShaderRD canvas_shader;
+		RID default_version;
+		RID default_version_rd_shader;
+		RID default_version_rd_shader_light;
+		RID quad_index_buffer;
+		RID quad_index_array;
+		PipelineVariants pipeline_variants;
+
+		// default_skeleton uniform set
+		RID default_skeleton_uniform_buffer;
+		RID default_skeleton_texture_buffer;
+
+		ShaderCompilerRD compiler;
+	} shader;
+
+	struct ShaderData : public RasterizerStorageRD::ShaderData {
+
+		enum BlendMode { //used internally
+			BLEND_MODE_MIX,
+			BLEND_MODE_ADD,
+			BLEND_MODE_SUB,
+			BLEND_MODE_MUL,
+			BLEND_MODE_PMALPHA,
+			BLEND_MODE_DISABLED,
+		};
+
+		enum LightMode {
+			LIGHT_MODE_NORMAL,
+			LIGHT_MODE_UNSHADED,
+			LIGHT_MODE_LIGHT_ONLY
+		};
+
+		bool valid;
+		RID version;
+		PipelineVariants pipeline_variants;
+		String path;
+
+		Map<StringName, ShaderLanguage::ShaderNode::Uniform> uniforms;
+		Vector<ShaderCompilerRD::GeneratedCode::Texture> texture_uniforms;
+
+		Vector<uint32_t> ubo_offsets;
+		uint32_t ubo_size;
+
+		String code;
+		Map<StringName, RID> default_texture_params;
+
+		bool uses_screen_texture;
+		bool uses_material_samplers;
+
+		virtual void set_code(const String &p_Code);
+		virtual void set_default_texture_param(const StringName &p_name, RID p_texture);
+		virtual void get_param_list(List<PropertyInfo> *p_param_list) const;
+		virtual bool is_param_texture(const StringName &p_param) const;
+		virtual bool is_animated() const;
+		virtual bool casts_shadows() const;
+		virtual Variant get_default_parameter(const StringName &p_parameter) const;
+		ShaderData();
+		virtual ~ShaderData();
+	};
+
+	RasterizerStorageRD::ShaderData *_create_shader_func();
+	static RasterizerStorageRD::ShaderData *_create_shader_funcs() {
+		return static_cast<RasterizerCanvasRD *>(singleton)->_create_shader_func();
+	}
+
+	struct MaterialData : public RasterizerStorageRD::MaterialData {
+		uint64_t last_frame;
+		ShaderData *shader_data;
+		RID uniform_buffer;
+		RID uniform_set;
+		Vector<RID> texture_cache;
+		Vector<uint8_t> ubo_data;
+
+		virtual void set_render_priority(int p_priority) {}
+		virtual void set_next_pass(RID p_pass) {}
+		virtual void update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty);
+		virtual ~MaterialData();
+	};
+
+	RasterizerStorageRD::MaterialData *_create_material_func(ShaderData *p_shader);
+	static RasterizerStorageRD::MaterialData *_create_material_funcs(RasterizerStorageRD::ShaderData *p_shader) {
+		return static_cast<RasterizerCanvasRD *>(singleton)->_create_material_func(static_cast<ShaderData *>(p_shader));
+	}
+
+	/**************************/
+	/**** TEXTURE BINDINGS ****/
+	/**************************/
+
+	// bindings used to render commands,
+	// cached for performance.
+
+	struct TextureBindingKey {
+		RID texture;
+		RID normalmap;
+		RID specular;
+		RID multimesh;
+		RS::CanvasItemTextureFilter texture_filter;
+		RS::CanvasItemTextureRepeat texture_repeat;
+		bool operator==(const TextureBindingKey &p_key) const {
+			return texture == p_key.texture && normalmap == p_key.normalmap && specular == p_key.specular && multimesh == p_key.specular && texture_filter == p_key.texture_filter && texture_repeat == p_key.texture_repeat;
+		}
+	};
+
+	struct TextureBindingKeyHasher {
+		static _FORCE_INLINE_ uint32_t hash(const TextureBindingKey &p_key) {
+			uint32_t hash = hash_djb2_one_64(p_key.texture.get_id());
+			hash = hash_djb2_one_64(p_key.normalmap.get_id(), hash);
+			hash = hash_djb2_one_64(p_key.specular.get_id(), hash);
+			hash = hash_djb2_one_64(p_key.multimesh.get_id(), hash);
+			hash = hash_djb2_one_32(uint32_t(p_key.texture_filter) << 16 | uint32_t(p_key.texture_repeat), hash);
+			return hash;
+		}
+	};
+
+	struct TextureBinding {
+		TextureBindingID id;
+		TextureBindingKey key;
+		SelfList<TextureBinding> to_dispose;
+		uint32_t reference_count;
+		RID uniform_set;
+		TextureBinding() :
+				to_dispose(this) {
+			reference_count = 0;
+		}
+	};
+
+	struct {
+		SelfList<TextureBinding>::List to_dispose_list;
+
+		TextureBindingID id_generator;
+		HashMap<TextureBindingKey, TextureBindingID, TextureBindingKeyHasher> texture_key_bindings;
+		HashMap<TextureBindingID, TextureBinding *> texture_bindings;
+
+		TextureBindingID default_empty;
+	} bindings;
+
+	RID _create_texture_binding(RID p_texture, RID p_normalmap, RID p_specular, RenderingServer::CanvasItemTextureFilter p_filter, RenderingServer::CanvasItemTextureRepeat p_repeat, RID p_multimesh);
+	void _dispose_bindings();
+
+	struct {
+		RS::CanvasItemTextureFilter default_filter;
+		RS::CanvasItemTextureRepeat default_repeat;
+	} default_samplers;
+
+	/******************/
+	/**** POLYGONS ****/
+	/******************/
+
+	struct PolygonBuffers {
+		RD::VertexFormatID vertex_format_id;
+		RID vertex_buffer;
+		RID vertex_array;
+		RID index_buffer;
+		RID indices;
+	};
+
+	struct {
+		HashMap<PolygonID, PolygonBuffers> polygons;
+		PolygonID last_id;
+	} polygon_buffers;
+
+	/********************/
+	/**** PRIMITIVES ****/
+	/********************/
+
+	struct {
+		RID index_array[4];
+	} primitive_arrays;
+
+	/*******************/
+	/**** MATERIALS ****/
+	/*******************/
+
+	/******************/
+	/**** LIGHTING ****/
+	/******************/
+
+	struct CanvasLight {
+
+		RID texture;
+		struct {
+			int size;
+			RID texture;
+			RID depth;
+			RID fb;
+		} shadow;
+	};
+
+	RID_Owner<CanvasLight> canvas_light_owner;
+
+	struct ShadowRenderPushConstant {
+		float projection[16];
+		float modelview[8];
+		float direction[2];
+		float pad[2];
+	};
+
+	struct OccluderPolygon {
+
+		RS::CanvasOccluderPolygonCullMode cull_mode;
+		int point_count;
+		RID vertex_buffer;
+		RID vertex_array;
+		RID index_buffer;
+		RID index_array;
+	};
+
+	struct LightUniform {
+		float matrix[8]; //light to texture coordinate matrix
+		float shadow_matrix[8]; //light to shadow coordinate matrix
+		float color[4];
+		float shadow_color[4];
+		float position[2];
+		uint32_t flags; //index to light texture
+		float height;
+		float shadow_pixel_size;
+		float pad[3];
+	};
+
+	RID_Owner<OccluderPolygon> occluder_polygon_owner;
+
+	struct {
+		CanvasOcclusionShaderRD shader;
+		RID shader_version;
+		RID render_pipelines[3];
+		RD::VertexFormatID vertex_format;
+		RD::FramebufferFormatID framebuffer_format;
+	} shadow_render;
+
+	/***************/
+	/**** STATE ****/
+	/***************/
+
+	//state that does not vary across rendering all items
+
+	struct ItemStateData : public Item::CustomData {
+
+		struct LightCache {
+			uint64_t light_version;
+			Light *light;
+		};
+
+		LightCache light_cache[DEFAULT_MAX_LIGHTS_PER_ITEM];
+		uint32_t light_cache_count;
+		RID state_uniform_set_with_light;
+		RID state_uniform_set;
+		ItemStateData() {
+
+			for (int i = 0; i < DEFAULT_MAX_LIGHTS_PER_ITEM; i++) {
+				light_cache[i].light_version = 0;
+				light_cache[i].light = NULL;
+			}
+			light_cache_count = 0xFFFFFFFF;
+		}
+
+		~ItemStateData() {
+			if (state_uniform_set_with_light.is_valid() && RD::get_singleton()->uniform_set_is_valid(state_uniform_set_with_light)) {
+				RD::get_singleton()->free(state_uniform_set_with_light);
+			}
+			if (state_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(state_uniform_set)) {
+				RD::get_singleton()->free(state_uniform_set);
+			}
+		}
+	};
+
+	struct State {
+
+		//state buffer
+		struct Buffer {
+			float canvas_transform[16];
+			float screen_transform[16];
+			float canvas_normal_transform[16];
+			float canvas_modulate[4];
+			float screen_pixel_size[2];
+			float time;
+			float pad;
+
+			//uint32_t light_count;
+			//uint32_t pad[3];
+		};
+
+		LightUniform *light_uniforms;
+
+		RID lights_uniform_buffer;
+		RID canvas_state_buffer;
+		RID shadow_sampler;
+
+		uint32_t max_lights_per_render;
+		uint32_t max_lights_per_item;
+
+		double time;
+	} state;
+
+	struct PushConstant {
+		float world[6];
+		uint32_t flags;
+		uint32_t specular_shininess;
+		union {
+			//rect
+			struct {
+				float modulation[4];
+				float ninepatch_margins[4];
+				float dst_rect[4];
+				float src_rect[4];
+				float pad[2];
+			};
+			//primitive
+			struct {
+				float points[6]; // vec2 points[3]
+				float uvs[6]; // vec2 points[3]
+				uint32_t colors[6]; // colors encoded as half
+			};
+		};
+		float color_texture_pixel_size[2];
+		uint32_t lights[4];
+	};
+
+	struct SkeletonUniform {
+		float skeleton_transform[16];
+		float skeleton_inverse[16];
+	};
+
+	Item *items[MAX_RENDER_ITEMS];
+
+	Size2i _bind_texture_binding(TextureBindingID p_binding, RenderingDevice::DrawListID p_draw_list, uint32_t &flags);
+	void _render_item(RenderingDevice::DrawListID p_draw_list, const Item *p_item, RenderingDevice::FramebufferFormatID p_framebuffer_format, const Transform2D &p_canvas_transform_inverse, Item *&current_clip, Light *p_lights, PipelineVariants *p_pipeline_variants);
+	void _render_items(RID p_to_render_target, int p_item_count, const Transform2D &p_canvas_transform_inverse, Light *p_lights, RID p_screen_uniform_set);
+
+	_FORCE_INLINE_ void _update_transform_2d_to_mat2x4(const Transform2D &p_transform, float *p_mat2x4);
+	_FORCE_INLINE_ void _update_transform_2d_to_mat2x3(const Transform2D &p_transform, float *p_mat2x3);
+
+	_FORCE_INLINE_ void _update_transform_2d_to_mat4(const Transform2D &p_transform, float *p_mat4);
+	_FORCE_INLINE_ void _update_transform_to_mat4(const Transform &p_transform, float *p_mat4);
+
+	_FORCE_INLINE_ void _update_specular_shininess(const Color &p_transform, uint32_t *r_ss);
+
+public:
+	TextureBindingID request_texture_binding(RID p_texture, RID p_normalmap, RID p_specular, RS::CanvasItemTextureFilter p_filter, RS::CanvasItemTextureRepeat p_repeat, RID p_multimesh);
+	void free_texture_binding(TextureBindingID p_binding);
+
+	PolygonID request_polygon(const Vector<int> &p_indices, const Vector<Point2> &p_points, const Vector<Color> &p_colors, const Vector<Point2> &p_uvs = Vector<Point2>(), const Vector<int> &p_bones = Vector<int>(), const Vector<float> &p_weights = Vector<float>());
+	void free_polygon(PolygonID p_polygon);
+
+	RID light_create();
+	void light_set_texture(RID p_rid, RID p_texture);
+	void light_set_use_shadow(RID p_rid, bool p_enable, int p_resolution);
+	void light_update_shadow(RID p_rid, const Transform2D &p_light_xform, int p_light_mask, float p_near, float p_far, LightOccluderInstance *p_occluders);
+
+	RID occluder_polygon_create();
+	void occluder_polygon_set_shape_as_lines(RID p_occluder, const Vector<Vector2> &p_lines);
+	void occluder_polygon_set_cull_mode(RID p_occluder, RS::CanvasOccluderPolygonCullMode p_mode);
+
+	void canvas_render_items(RID p_to_render_target, Item *p_item_list, const Color &p_modulate, Light *p_light_list, const Transform2D &p_canvas_transform);
+
+	void canvas_debug_viewport_shadows(Light *p_lights_with_shadow){};
+
+	void draw_window_margins(int *p_margins, RID *p_margin_textures) {}
+
+	void set_time(double p_time);
+	void update();
+	bool free(RID p_rid);
+	RasterizerCanvasRD(RasterizerStorageRD *p_storage);
+	~RasterizerCanvasRD();
+};
+
+#endif // RASTERIZER_CANVAS_RD_H

servers/rendering/rasterizer_rd/rasterizer_effects_rd.h

@@ -0,0 +1,457 @@
+/*************************************************************************/
+/*  rasterizer_effects_rd.h                                              */
+/*************************************************************************/
+/*                       This file is part of:                           */
+/*                           GODOT ENGINE                                */
+/*                      https://godotengine.org                          */
+/*************************************************************************/
+/* Copyright (c) 2007-2020 Juan Linietsky, Ariel Manzur.                 */
+/* Copyright (c) 2014-2020 Godot Engine contributors (cf. AUTHORS.md).   */
+/*                                                                       */
+/* Permission is hereby granted, free of charge, to any person obtaining */
+/* a copy of this software and associated documentation files (the       */
+/* "Software"), to deal in the Software without restriction, including   */
+/* without limitation the rights to use, copy, modify, merge, publish,   */
+/* distribute, sublicense, and/or sell copies of the Software, and to    */
+/* permit persons to whom the Software is furnished to do so, subject to */
+/* the following conditions:                                             */
+/*                                                                       */
+/* The above copyright notice and this permission notice shall be        */
+/* included in all copies or substantial portions of the Software.       */
+/*                                                                       */
+/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,       */
+/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF    */
+/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
+/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY  */
+/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,  */
+/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE     */
+/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                */
+/*************************************************************************/
+
+#ifndef RASTERIZER_EFFECTS_RD_H
+#define RASTERIZER_EFFECTS_RD_H
+
+#include "core/math/camera_matrix.h"
+#include "servers/rendering/rasterizer_rd/render_pipeline_vertex_format_cache_rd.h"
+#include "servers/rendering/rasterizer_rd/shaders/blur.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/bokeh_dof.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/copy.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/cubemap_downsampler.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/cubemap_filter.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/cubemap_roughness.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/luminance_reduce.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/roughness_limiter.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/ssao.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/ssao_blur.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/ssao_minify.glsl.gen.h"
+#include "servers/rendering/rasterizer_rd/shaders/tonemap.glsl.gen.h"
+
+#include "servers/rendering_server.h"
+
+class RasterizerEffectsRD {
+
+	enum BlurMode {
+		BLUR_MODE_GAUSSIAN_BLUR,
+		BLUR_MODE_GAUSSIAN_GLOW,
+		BLUR_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE,
+		BLUR_MODE_DOF_NEAR_LOW,
+		BLUR_MODE_DOF_NEAR_MEDIUM,
+		BLUR_MODE_DOF_NEAR_HIGH,
+		BLUR_MODE_DOF_NEAR_MERGE_LOW,
+		BLUR_MODE_DOF_NEAR_MERGE_MEDIUM,
+		BLUR_MODE_DOF_NEAR_MERGE_HIGH,
+		BLUR_MODE_DOF_FAR_LOW,
+		BLUR_MODE_DOF_FAR_MEDIUM,
+		BLUR_MODE_DOF_FAR_HIGH,
+		BLUR_MODE_SSAO_MERGE,
+		BLUR_MODE_SIMPLY_COPY,
+		BLUR_MODE_MIPMAP,
+		BLUR_MODE_MAX,
+
+	};
+
+	enum {
+		BLUR_FLAG_HORIZONTAL = (1 << 0),
+		BLUR_FLAG_USE_BLUR_SECTION = (1 << 1),
+		BLUR_FLAG_USE_ORTHOGONAL_PROJECTION = (1 << 2),
+		BLUR_FLAG_DOF_NEAR_FIRST_TAP = (1 << 3),
+		BLUR_FLAG_GLOW_FIRST_PASS = (1 << 4),
+		BLUR_FLAG_FLIP_Y = (1 << 5),
+		BLUR_COPY_FORCE_LUMINANCE = (1 << 6)
+	};
+
+	struct BlurPushConstant {
+		float section[4];
+		float pixel_size[2];
+		uint32_t flags;
+		uint32_t pad;
+		//glow
+		float glow_strength;
+		float glow_bloom;
+		float glow_hdr_threshold;
+		float glow_hdr_scale;
+		float glow_exposure;
+		float glow_white;
+		float glow_luminance_cap;
+		float glow_auto_exposure_grey;
+		//dof
+		float dof_begin;
+		float dof_end;
+		float dof_radius;
+		float dof_pad;
+
+		float dof_dir[2];
+		float camera_z_far;
+		float camera_z_near;
+
+		float ssao_color[4];
+	};
+
+	struct Blur {
+		BlurPushConstant push_constant;
+		BlurShaderRD shader;
+		RID shader_version;
+		RenderPipelineVertexFormatCacheRD pipelines[BLUR_MODE_MAX];
+
+	} blur;
+
+	struct CubemapRoughnessPushConstant {
+		uint32_t face_id;
+		uint32_t sample_count;
+		float roughness;
+		uint32_t use_direct_write;
+		float face_size;
+		float pad[3];
+	};
+
+	struct CubemapRoughness {
+
+		CubemapRoughnessPushConstant push_constant;
+		CubemapRoughnessShaderRD shader;
+		RID shader_version;
+		RID pipeline;
+	} roughness;
+
+	enum TonemapMode {
+		TONEMAP_MODE_NORMAL,
+		TONEMAP_MODE_BICUBIC_GLOW_FILTER,
+		TONEMAP_MODE_MAX
+	};
+
+	struct TonemapPushConstant {
+		float bcs[3];
+		uint32_t use_bcs;
+
+		uint32_t use_glow;
+		uint32_t use_auto_exposure;
+		uint32_t use_color_correction;
+		uint32_t tonemapper;
+
+		uint32_t glow_texture_size[2];
+
+		float glow_intensity;
+		uint32_t glow_level_flags;
+		uint32_t glow_mode;
+
+		float exposure;
+		float white;
+		float auto_exposure_grey;
+	};
+
+	struct Tonemap {
+
+		TonemapPushConstant push_constant;
+		TonemapShaderRD shader;
+		RID shader_version;
+		RenderPipelineVertexFormatCacheRD pipelines[TONEMAP_MODE_MAX];
+	} tonemap;
+
+	enum LuminanceReduceMode {
+		LUMINANCE_REDUCE_READ,
+		LUMINANCE_REDUCE,
+		LUMINANCE_REDUCE_WRITE,
+		LUMINANCE_REDUCE_MAX
+	};
+
+	struct LuminanceReducePushConstant {
+		int32_t source_size[2];
+		float max_luminance;
+		float min_luminance;
+		float exposure_adjust;
+		float pad[3];
+	};
+
+	struct LuminanceReduce {
+
+		LuminanceReducePushConstant push_constant;
+		LuminanceReduceShaderRD shader;
+		RID shader_version;
+		RID pipelines[LUMINANCE_REDUCE_MAX];
+	} luminance_reduce;
+
+	struct CopyToDPPushConstant {
+		float bias;
+		float z_far;
+		float z_near;
+		uint32_t z_flip;
+	};
+
+	enum CopyMode {
+		COPY_MODE_CUBE_TO_DP,
+		COPY_MODE_MAX
+	};
+
+	struct Copy {
+
+		CopyShaderRD shader;
+		RID shader_version;
+		RenderPipelineVertexFormatCacheRD pipelines[COPY_MODE_MAX];
+	} copy;
+
+	struct BokehPushConstant {
+		uint32_t size[2];
+		float z_far;
+		float z_near;
+
+		uint32_t orthogonal;
+		float blur_size;
+		float blur_scale;
+		uint32_t steps;
+
+		uint32_t blur_near_active;
+		float blur_near_begin;
+		float blur_near_end;
+		uint32_t blur_far_active;
+
+		float blur_far_begin;
+		float blur_far_end;
+		uint32_t second_pass;
+		uint32_t half_size;
+
+		uint32_t use_jitter;
+		float jitter_seed;
+		uint32_t pad[2];
+	};
+
+	enum BokehMode {
+		BOKEH_GEN_BLUR_SIZE,
+		BOKEH_GEN_BOKEH_BOX,
+		BOKEH_GEN_BOKEH_HEXAGONAL,
+		BOKEH_GEN_BOKEH_CIRCULAR,
+		BOKEH_COMPOSITE,
+		BOKEH_MAX
+	};
+
+	struct Bokeh {
+
+		BokehPushConstant push_constant;
+		BokehDofShaderRD shader;
+		RID shader_version;
+		RID pipelines[BOKEH_MAX];
+	} bokeh;
+
+	enum SSAOMode {
+		SSAO_MINIFY_FIRST,
+		SSAO_MINIFY_MIPMAP,
+		SSAO_GATHER_LOW,
+		SSAO_GATHER_MEDIUM,
+		SSAO_GATHER_HIGH,
+		SSAO_GATHER_ULTRA,
+		SSAO_GATHER_LOW_HALF,
+		SSAO_GATHER_MEDIUM_HALF,
+		SSAO_GATHER_HIGH_HALF,
+		SSAO_GATHER_ULTRA_HALF,
+		SSAO_BLUR_PASS,
+		SSAO_BLUR_PASS_HALF,
+		SSAO_BLUR_UPSCALE,
+		SSAO_MAX
+	};
+
+	struct SSAOMinifyPushConstant {
+		float pixel_size[2];
+		float z_far;
+		float z_near;
+		int32_t source_size[2];
+		uint32_t orthogonal;
+		uint32_t pad;
+	};
+
+	struct SSAOGatherPushConstant {
+		int32_t screen_size[2];
+		float z_far;
+		float z_near;
+
+		uint32_t orthogonal;
+		float intensity_div_r6;
+		float radius;
+		float bias;
+
+		float proj_info[4];
+		float pixel_size[2];
+		float proj_scale;
+		uint32_t pad;
+	};
+
+	struct SSAOBlurPushConstant {
+		float edge_sharpness;
+		int32_t filter_scale;
+		float z_far;
+		float z_near;
+		uint32_t orthogonal;
+		uint32_t pad[3];
+		int32_t axis[2];
+		int32_t screen_size[2];
+	};
+
+	struct SSAO {
+
+		SSAOMinifyPushConstant minify_push_constant;
+		SsaoMinifyShaderRD minify_shader;
+		RID minify_shader_version;
+
+		SSAOGatherPushConstant gather_push_constant;
+		SsaoShaderRD gather_shader;
+		RID gather_shader_version;
+
+		SSAOBlurPushConstant blur_push_constant;
+		SsaoBlurShaderRD blur_shader;
+		RID blur_shader_version;
+
+		RID pipelines[SSAO_MAX];
+	} ssao;
+
+	struct RoughnessLimiterPushConstant {
+		int32_t screen_size[2];
+		float curve;
+		uint32_t pad;
+	};
+
+	struct RoughnessLimiter {
+
+		RoughnessLimiterPushConstant push_constant;
+		RoughnessLimiterShaderRD shader;
+		RID shader_version;
+		RID pipeline;
+
+	} roughness_limiter;
+
+	struct CubemapDownsamplerPushConstant {
+		uint32_t face_size;
+		float pad[3];
+	};
+
+	struct CubemapDownsampler {
+
+		CubemapDownsamplerPushConstant push_constant;
+		CubemapDownsamplerShaderRD shader;
+		RID shader_version;
+		RID pipeline;
+
+	} cubemap_downsampler;
+
+	enum CubemapFilterMode {
+		FILTER_MODE_HIGH_QUALITY,
+		FILTER_MODE_LOW_QUALITY,
+		FILTER_MODE_HIGH_QUALITY_ARRAY,
+		FILTER_MODE_LOW_QUALITY_ARRAY,
+		FILTER_MODE_MAX,
+	};
+
+	struct CubemapFilter {
+
+		CubemapFilterShaderRD shader;
+		RID shader_version;
+		RID pipelines[FILTER_MODE_MAX];
+		RID uniform_set;
+		RID image_uniform_set;
+		RID coefficient_buffer;
+		bool use_high_quality;
+
+	} filter;
+
+	struct SkyPushConstant {
+		float orientation[12];
+		float proj[4];
+		float position[3];
+		float multiplier;
+		float time;
+		float pad[3];
+	};
+
+	RID default_sampler;
+	RID default_mipmap_sampler;
+	RID index_buffer;
+	RID index_array;
+
+	Map<RID, RID> texture_to_uniform_set_cache;
+
+	Map<RID, RID> image_to_uniform_set_cache;
+	Map<RID, RID> texture_to_compute_uniform_set_cache;
+
+	RID _get_uniform_set_from_image(RID p_texture);
+	RID _get_uniform_set_from_texture(RID p_texture, bool p_use_mipmaps = false);
+	RID _get_compute_uniform_set_from_texture(RID p_texture, bool p_use_mipmaps = false);
+
+public:
+	//TODO must re-do most of the shaders in compute
+
+	void region_copy(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2 &p_region);
+	void copy_to_rect(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2 &p_rect, bool p_flip_y = false, bool p_force_luminance = false);
+	void gaussian_blur(RID p_source_rd_texture, RID p_framebuffer_half, RID p_rd_texture_half, RID p_dest_framebuffer, const Vector2 &p_pixel_size, const Rect2 &p_region);
+	void gaussian_glow(RID p_source_rd_texture, RID p_framebuffer_half, RID p_rd_texture_half, RID p_dest_framebuffer, const Vector2 &p_pixel_size, float p_strength = 1.0, bool p_first_pass = false, float p_luminance_cap = 16.0, float p_exposure = 1.0, float p_bloom = 0.0, float p_hdr_bleed_treshold = 1.0, float p_hdr_bleed_scale = 1.0, RID p_auto_exposure = RID(), float p_auto_exposure_grey = 1.0);
+
+	void cubemap_roughness(RID p_source_rd_texture, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size);
+	void make_mipmap(RID p_source_rd_texture, RID p_framebuffer_half, const Vector2 &p_pixel_size);
+	void copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2 &p_rect, float p_z_near, float p_z_far, float p_bias, bool p_dp_flip);
+	void luminance_reduction(RID p_source_texture, const Size2i p_source_size, const Vector<RID> p_reduce, RID p_prev_luminance, float p_min_luminance, float p_max_luminance, float p_adjust, bool p_set = false);
+	void bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i &p_base_texture_size, RID p_secondary_texture, RID p_bokeh_texture1, RID p_bokeh_texture2, bool p_dof_far, float p_dof_far_begin, float p_dof_far_size, bool p_dof_near, float p_dof_near_begin, float p_dof_near_size, float p_bokeh_size, RS::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, bool p_use_jitter, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal);
+
+	struct TonemapSettings {
+
+		bool use_glow = false;
+		enum GlowMode {
+			GLOW_MODE_ADD,
+			GLOW_MODE_SCREEN,
+			GLOW_MODE_SOFTLIGHT,
+			GLOW_MODE_REPLACE,
+			GLOW_MODE_MIX
+		};
+
+		GlowMode glow_mode = GLOW_MODE_ADD;
+		float glow_intensity = 1.0;
+		uint32_t glow_level_flags = 0;
+		Vector2i glow_texture_size;
+		bool glow_use_bicubic_upscale = false;
+		RID glow_texture;
+
+		RS::EnvironmentToneMapper tonemap_mode = RS::ENV_TONE_MAPPER_LINEAR;
+		float exposure = 1.0;
+		float white = 1.0;
+
+		bool use_auto_exposure = false;
+		float auto_exposure_grey = 0.5;
+		RID exposure_texture;
+
+		bool use_bcs = false;
+		float brightness = 1.0;
+		float contrast = 1.0;
+		float saturation = 1.0;
+
+		bool use_color_correction = false;
+		RID color_correction_texture;
+	};
+
+	void tonemapper(RID p_source_color, RID p_dst_framebuffer, const TonemapSettings &p_settings);
+
+	void generate_ssao(RID p_depth_buffer, RID p_normal_buffer, const Size2i &p_depth_buffer_size, RID p_depth_mipmaps_texture, const Vector<RID> &depth_mipmaps, RID p_ao1, bool p_half_size, RID p_ao2, RID p_upscale_buffer, float p_intensity, float p_radius, float p_bias, const CameraMatrix &p_projection, RS::EnvironmentSSAOQuality p_quality, RS::EnvironmentSSAOBlur p_blur, float p_edge_sharpness);
+
+	void roughness_limit(RID p_source_normal, RID p_roughness, const Size2i &p_size, float p_curve);
+	void cubemap_downsample(RID p_source_cubemap, RID p_dest_cubemap, const Size2i &p_size);
+	void cubemap_filter(RID p_source_cubemap, Vector<RID> p_dest_cubemap, bool p_use_array);
+	void render_sky(RD::DrawListID p_list, float p_time, RID p_fb, RID p_samplers, RID p_lights, RenderPipelineVertexFormatCacheRD *p_pipeline, RID p_uniform_set, RID p_texture_set, const CameraMatrix &p_camera, const Basis &p_orientation, float p_multiplier, const Vector3 &p_position);
+
+	RasterizerEffectsRD();
+	~RasterizerEffectsRD();
+};
+
+#endif // !RASTERIZER_EFFECTS_RD_H

servers/rendering/rasterizer_rd/rasterizer_rd.cpp

@@ -0,0 +1,177 @@
+/*************************************************************************/
+/*  rasterizer_rd.cpp                                                    */
+/*************************************************************************/
+/*                       This file is part of:                           */
+/*                           GODOT ENGINE                                */
+/*                      https://godotengine.org                          */
+/*************************************************************************/
+/* Copyright (c) 2007-2020 Juan Linietsky, Ariel Manzur.                 */
+/* Copyright (c) 2014-2020 Godot Engine contributors (cf. AUTHORS.md).   */
+/*                                                                       */
+/* Permission is hereby granted, free of charge, to any person obtaining */
+/* a copy of this software and associated documentation files (the       */
+/* "Software"), to deal in the Software without restriction, including   */
+/* without limitation the rights to use, copy, modify, merge, publish,   */
+/* distribute, sublicense, and/or sell copies of the Software, and to    */
+/* permit persons to whom the Software is furnished to do so, subject to */
+/* the following conditions:                                             */
+/*                                                                       */
+/* The above copyright notice and this permission notice shall be        */
+/* included in all copies or substantial portions of the Software.       */
+/*                                                                       */
+/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,       */
+/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF    */
+/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
+/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY  */
+/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,  */
+/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE     */
+/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                */
+/*************************************************************************/
+
+#include "rasterizer_rd.h"
+
+void RasterizerRD::prepare_for_blitting_render_targets() {
+	RD::get_singleton()->prepare_screen_for_drawing();
+}
+
+void RasterizerRD::blit_render_targets_to_screen(DisplayServer::WindowID p_screen, const BlitToScreen *p_render_targets, int p_amount) {
+
+	RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin_for_screen(p_screen);
+
+	for (int i = 0; i < p_amount; i++) {
+		RID texture = storage->render_target_get_texture(p_render_targets[i].render_target);
+		ERR_CONTINUE(texture.is_null());
+		RID rd_texture = storage->texture_get_rd_texture(texture);
+		ERR_CONTINUE(rd_texture.is_null());
+		if (!render_target_descriptors.has(rd_texture) || !RD::get_singleton()->uniform_set_is_valid(render_target_descriptors[rd_texture])) {
+			Vector<RD::Uniform> uniforms;
+			RD::Uniform u;
+			u.type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE;
+			u.binding = 0;
+			u.ids.push_back(copy_viewports_sampler);
+			u.ids.push_back(rd_texture);
+			uniforms.push_back(u);
+			RID uniform_set = RD::get_singleton()->uniform_set_create(uniforms, copy_viewports_rd_shader, 0);
+
+			render_target_descriptors[rd_texture] = uniform_set;
+		}
+
+		Size2 screen_size(RD::get_singleton()->screen_get_width(p_screen), RD::get_singleton()->screen_get_height(p_screen));
+
+		RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, copy_viewports_rd_pipeline);
+		RD::get_singleton()->draw_list_bind_index_array(draw_list, copy_viewports_rd_array);
+		RD::get_singleton()->draw_list_bind_uniform_set(draw_list, render_target_descriptors[rd_texture], 0);
+
+		float push_constant[4] = {
+			p_render_targets[i].rect.position.x / screen_size.width,
+			p_render_targets[i].rect.position.y / screen_size.height,
+			p_render_targets[i].rect.size.width / screen_size.width,
+			p_render_targets[i].rect.size.height / screen_size.height,
+		};
+		RD::get_singleton()->draw_list_set_push_constant(draw_list, push_constant, 4 * sizeof(float));
+		RD::get_singleton()->draw_list_draw(draw_list, true);
+	}
+
+	RD::get_singleton()->draw_list_end();
+}
+
+void RasterizerRD::begin_frame(double frame_step) {
+	frame++;
+	time += frame_step;
+	canvas->set_time(time);
+	scene->set_time(time, frame_step);
+}
+
+void RasterizerRD::end_frame(bool p_swap_buffers) {
+
+#ifndef _MSC_VER
+#warning TODO: likely passa bool to swap buffers to avoid display?
+#endif
+	RD::get_singleton()->swap_buffers(); //probably should pass some bool to avoid display?
+}
+
+void RasterizerRD::initialize() {
+
+	{ //create framebuffer copy shader
+		RenderingDevice::ShaderStageData vert;
+		vert.shader_stage = RenderingDevice::SHADER_STAGE_VERTEX;
+		vert.spir_v = RenderingDevice::get_singleton()->shader_compile_from_source(RenderingDevice::SHADER_STAGE_VERTEX,
+				"#version 450\n"
+				"layout(push_constant, binding = 0, std140) uniform Pos { vec4 dst_rect; } pos;\n"
+				"layout(location =0) out vec2 uv;\n"
+				"void main() { \n"
+				" vec2 base_arr[4] = vec2[](vec2(0.0,0.0),vec2(0.0,1.0),vec2(1.0,1.0),vec2(1.0,0.0));\n"
+				" uv = base_arr[gl_VertexIndex];\n"
+				" vec2 vtx = pos.dst_rect.xy+uv*pos.dst_rect.zw;\n"
+				" gl_Position = vec4(vtx * 2.0 - 1.0,0.0,1.0);\n"
+				"}\n");
+
+		RenderingDevice::ShaderStageData frag;
+		frag.shader_stage = RenderingDevice::SHADER_STAGE_FRAGMENT;
+		frag.spir_v = RenderingDevice::get_singleton()->shader_compile_from_source(RenderingDevice::SHADER_STAGE_FRAGMENT,
+				"#version 450\n"
+				"layout (location = 0) in vec2 uv;\n"
+				"layout (location = 0) out vec4 color;\n"
+				"layout (binding = 0) uniform sampler2D src_rt;\n"
+				"void main() { color=texture(src_rt,uv); }\n");
+
+		Vector<RenderingDevice::ShaderStageData> source;
+		source.push_back(vert);
+		source.push_back(frag);
+		String error;
+		copy_viewports_rd_shader = RD::get_singleton()->shader_create(source);
+		if (!copy_viewports_rd_shader.is_valid()) {
+			print_line("Failed compilation: " + error);
+		}
+	}
+
+	{ //create index array for copy shader
+		Vector<uint8_t> pv;
+		pv.resize(6 * 4);
+		{
+			uint8_t *w = pv.ptrw();
+			int *p32 = (int *)w;
+			p32[0] = 0;
+			p32[1] = 1;
+			p32[2] = 2;
+			p32[3] = 0;
+			p32[4] = 2;
+			p32[5] = 3;
+		}
+		copy_viewports_rd_index_buffer = RD::get_singleton()->index_buffer_create(6, RenderingDevice::INDEX_BUFFER_FORMAT_UINT32, pv);
+		copy_viewports_rd_array = RD::get_singleton()->index_array_create(copy_viewports_rd_index_buffer, 0, 6);
+	}
+
+	{ //pipeline
+		copy_viewports_rd_pipeline = RD::get_singleton()->render_pipeline_create(copy_viewports_rd_shader, RD::get_singleton()->screen_get_framebuffer_format(), RD::INVALID_ID, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RenderingDevice::PipelineColorBlendState::create_disabled(), 0);
+	}
+	{ // sampler
+		copy_viewports_sampler = RD::get_singleton()->sampler_create(RD::SamplerState());
+	}
+}
+
+ThreadWorkPool RasterizerRD::thread_work_pool;
+uint32_t RasterizerRD::frame = 1;
+
+void RasterizerRD::finalize() {
+
+	thread_work_pool.finish();
+
+	memdelete(scene);
+	memdelete(canvas);
+	memdelete(storage);
+
+	//only need to erase these, the rest are erased by cascade
+	RD::get_singleton()->free(copy_viewports_rd_index_buffer);
+	RD::get_singleton()->free(copy_viewports_rd_shader);
+	RD::get_singleton()->free(copy_viewports_sampler);
+}
+
+RasterizerRD::RasterizerRD() {
+	thread_work_pool.init();
+	time = 0;
+
+	storage = memnew(RasterizerStorageRD);
+	canvas = memnew(RasterizerCanvasRD(storage));
+	scene = memnew(RasterizerSceneHighEndRD(storage));
+}

servers/rendering/rasterizer_rd/rasterizer_rd.h

@@ -0,0 +1,95 @@
+/*************************************************************************/
+/*  rasterizer_rd.h                                                      */
+/*************************************************************************/
+/*                       This file is part of:                           */
+/*                           GODOT ENGINE                                */
+/*                      https://godotengine.org                          */
+/*************************************************************************/
+/* Copyright (c) 2007-2020 Juan Linietsky, Ariel Manzur.                 */
+/* Copyright (c) 2014-2020 Godot Engine contributors (cf. AUTHORS.md).   */
+/*                                                                       */
+/* Permission is hereby granted, free of charge, to any person obtaining */
+/* a copy of this software and associated documentation files (the       */
+/* "Software"), to deal in the Software without restriction, including   */
+/* without limitation the rights to use, copy, modify, merge, publish,   */
+/* distribute, sublicense, and/or sell copies of the Software, and to    */
+/* permit persons to whom the Software is furnished to do so, subject to */
+/* the following conditions:                                             */
+/*                                                                       */
+/* The above copyright notice and this permission notice shall be        */
+/* included in all copies or substantial portions of the Software.       */
+/*                                                                       */
+/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,       */
+/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF    */
+/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
+/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY  */
+/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,  */
+/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE     */
+/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                */
+/*************************************************************************/
+
+#ifndef RASTERIZER_RD_H
+#define RASTERIZER_RD_H
+
+#include "core/os/os.h"
+#include "core/thread_work_pool.h"
+#include "servers/rendering/rasterizer.h"
+#include "servers/rendering/rasterizer_rd/rasterizer_canvas_rd.h"
+#include "servers/rendering/rasterizer_rd/rasterizer_scene_high_end_rd.h"
+#include "servers/rendering/rasterizer_rd/rasterizer_storage_rd.h"
+
+class RasterizerRD : public Rasterizer {
+protected:
+	RasterizerCanvasRD *canvas;
+	RasterizerStorageRD *storage;
+	RasterizerSceneHighEndRD *scene;
+
+	RID copy_viewports_rd_shader;
+	RID copy_viewports_rd_pipeline;
+	RID copy_viewports_rd_index_buffer;
+	RID copy_viewports_rd_array;
+	RID copy_viewports_sampler;
+
+	Map<RID, RID> render_target_descriptors;
+
+	double time;
+
+	static uint32_t frame;
+
+public:
+	RasterizerStorage *get_storage() { return storage; }
+	RasterizerCanvas *get_canvas() { return canvas; }
+	RasterizerScene *get_scene() { return scene; }
+
+	void set_boot_image(const Ref<Image> &p_image, const Color &p_color, bool p_scale, bool p_use_filter) {}
+
+	void initialize();
+	void begin_frame(double frame_step);
+	void prepare_for_blitting_render_targets();
+	void blit_render_targets_to_screen(DisplayServer::WindowID p_screen, const BlitToScreen *p_render_targets, int p_amount);
+
+	void end_frame(bool p_swap_buffers);
+	void finalize();
+
+	static _ALWAYS_INLINE_ uint64_t get_frame_number() { return frame; }
+
+	static Error is_viable() {
+		return OK;
+	}
+
+	static Rasterizer *_create_current() {
+		return memnew(RasterizerRD);
+	}
+
+	static void make_current() {
+		_create_func = _create_current;
+	}
+
+	virtual bool is_low_end() const { return false; }
+
+	static ThreadWorkPool thread_work_pool;
+
+	RasterizerRD();
+	~RasterizerRD() {}
+};
+#endif // RASTERIZER_RD_H

servers/rendering/rasterizer_rd/rasterizer_scene_high_end_rd.h

@@ -0,0 +1,585 @@
+/*************************************************************************/
+/*  rasterizer_scene_high_end_rd.h                                       */
+/*************************************************************************/
+/*                       This file is part of:                           */
+/*                           GODOT ENGINE                                */
+/*                      https://godotengine.org                          */
+/*************************************************************************/
+/* Copyright (c) 2007-2020 Juan Linietsky, Ariel Manzur.                 */
+/* Copyright (c) 2014-2020 Godot Engine contributors (cf. AUTHORS.md).   */
+/*                                                                       */
+/* Permission is hereby granted, free of charge, to any person obtaining */
+/* a copy of this software and associated documentation files (the       */
+/* "Software"), to deal in the Software without restriction, including   */
+/* without limitation the rights to use, copy, modify, merge, publish,   */
+/* distribute, sublicense, and/or sell copies of the Software, and to    */
+/* permit persons to whom the Software is furnished to do so, subject to */
+/* the following conditions:                                             */
+/*                                                                       */
+/* The above copyright notice and this permission notice shall be        */
+/* included in all copies or substantial portions of the Software.       */
+/*                                                                       */
+/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,       */
+/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF    */
+/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
+/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY  */
+/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,  */
+/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE     */
+/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                */
+/*************************************************************************/
+
+#ifndef RASTERIZER_SCENE_HIGHEND_RD_H
+#define RASTERIZER_SCENE_HIGHEND_RD_H
+
+#include "servers/rendering/rasterizer_rd/light_cluster_builder.h"
+#include "servers/rendering/rasterizer_rd/rasterizer_scene_rd.h"
+#include "servers/rendering/rasterizer_rd/rasterizer_storage_rd.h"
+#include "servers/rendering/rasterizer_rd/render_pipeline_vertex_format_cache_rd.h"
+#include "servers/rendering/rasterizer_rd/shaders/scene_high_end.glsl.gen.h"
+
+class RasterizerSceneHighEndRD : public RasterizerSceneRD {
+
+	enum {
+		SCENE_UNIFORM_SET = 0,
+		RADIANCE_UNIFORM_SET = 1,
+		VIEW_DEPENDANT_UNIFORM_SET = 2,
+		RENDER_BUFFERS_UNIFORM_SET = 3,
+		TRANSFORMS_UNIFORM_SET = 4,
+		MATERIAL_UNIFORM_SET = 5
+	};
+
+	/* Scene Shader */
+
+	enum ShaderVersion {
+		SHADER_VERSION_DEPTH_PASS,
+		SHADER_VERSION_DEPTH_PASS_DP,
+		SHADER_VERSION_DEPTH_PASS_WITH_NORMAL,
+		SHADER_VERSION_DEPTH_PASS_WITH_NORMAL_AND_ROUGHNESS,
+		SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL,
+		SHADER_VERSION_COLOR_PASS,
+		SHADER_VERSION_COLOR_PASS_WITH_SEPARATE_SPECULAR,
+		SHADER_VERSION_VCT_COLOR_PASS,
+		SHADER_VERSION_VCT_COLOR_PASS_WITH_SEPARATE_SPECULAR,
+		SHADER_VERSION_LIGHTMAP_COLOR_PASS,
+		SHADER_VERSION_LIGHTMAP_COLOR_PASS_WITH_SEPARATE_SPECULAR,
+		SHADER_VERSION_MAX
+	};
+
+	struct {
+		SceneHighEndShaderRD scene_shader;
+		ShaderCompilerRD compiler;
+	} shader;
+
+	RasterizerStorageRD *storage;
+
+	/* Material */
+
+	struct ShaderData : public RasterizerStorageRD::ShaderData {
+
+		enum BlendMode { //used internally
+			BLEND_MODE_MIX,
+			BLEND_MODE_ADD,
+			BLEND_MODE_SUB,
+			BLEND_MODE_MUL,
+		};
+
+		enum DepthDraw {
+			DEPTH_DRAW_DISABLED,
+			DEPTH_DRAW_OPAQUE,
+			DEPTH_DRAW_ALWAYS
+		};
+
+		enum DepthTest {
+			DEPTH_TEST_DISABLED,
+			DEPTH_TEST_ENABLED
+		};
+
+		enum Cull {
+			CULL_DISABLED,
+			CULL_FRONT,
+			CULL_BACK
+		};
+
+		enum CullVariant {
+			CULL_VARIANT_NORMAL,
+			CULL_VARIANT_REVERSED,
+			CULL_VARIANT_DOUBLE_SIDED,
+			CULL_VARIANT_MAX
+
+		};
+
+		bool valid;
+		RID version;
+		uint32_t vertex_input_mask;
+		RenderPipelineVertexFormatCacheRD pipelines[CULL_VARIANT_MAX][RS::PRIMITIVE_MAX][SHADER_VERSION_MAX];
+
+		String path;
+
+		Map<StringName, ShaderLanguage::ShaderNode::Uniform> uniforms;
+		Vector<ShaderCompilerRD::GeneratedCode::Texture> texture_uniforms;
+
+		Vector<uint32_t> ubo_offsets;
+		uint32_t ubo_size;
+
+		String code;
+		Map<StringName, RID> default_texture_params;
+
+		DepthDraw depth_draw;
+		DepthTest depth_test;
+
+		bool uses_point_size;
+		bool uses_alpha;
+		bool uses_blend_alpha;
+		bool uses_depth_pre_pass;
+		bool uses_discard;
+		bool uses_roughness;
+		bool uses_normal;
+
+		bool unshaded;
+		bool uses_vertex;
+		bool uses_sss;
+		bool uses_screen_texture;
+		bool uses_depth_texture;
+		bool uses_normal_texture;
+		bool uses_time;
+		bool writes_modelview_or_projection;
+		bool uses_world_coordinates;
+
+		uint64_t last_pass = 0;
+		uint32_t index = 0;
+
+		virtual void set_code(const String &p_Code);
+		virtual void set_default_texture_param(const StringName &p_name, RID p_texture);
+		virtual void get_param_list(List<PropertyInfo> *p_param_list) const;
+		virtual bool is_param_texture(const StringName &p_param) const;
+		virtual bool is_animated() const;
+		virtual bool casts_shadows() const;
+		virtual Variant get_default_parameter(const StringName &p_parameter) const;
+		ShaderData();
+		virtual ~ShaderData();
+	};
+
+	RasterizerStorageRD::ShaderData *_create_shader_func();
+	static RasterizerStorageRD::ShaderData *_create_shader_funcs() {
+		return static_cast<RasterizerSceneHighEndRD *>(singleton)->_create_shader_func();
+	}
+
+	struct MaterialData : public RasterizerStorageRD::MaterialData {
+		uint64_t last_frame;
+		ShaderData *shader_data;
+		RID uniform_buffer;
+		RID uniform_set;
+		Vector<RID> texture_cache;
+		Vector<uint8_t> ubo_data;
+		uint64_t last_pass = 0;
+		uint32_t index = 0;
+		RID next_pass;
+		uint8_t priority;
+		virtual void set_render_priority(int p_priority);
+		virtual void set_next_pass(RID p_pass);
+		virtual void update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty);
+		virtual ~MaterialData();
+	};
+
+	RasterizerStorageRD::MaterialData *_create_material_func(ShaderData *p_shader);
+	static RasterizerStorageRD::MaterialData *_create_material_funcs(RasterizerStorageRD::ShaderData *p_shader) {
+		return static_cast<RasterizerSceneHighEndRD *>(singleton)->_create_material_func(static_cast<ShaderData *>(p_shader));
+	}
+
+	/* Push Constant */
+
+	struct PushConstant {
+		uint32_t index;
+		uint32_t pad[3];
+	};
+
+	/* Framebuffer */
+
+	struct RenderBufferDataHighEnd : public RenderBufferData {
+		//for rendering, may be MSAAd
+		RID color;
+		RID depth;
+		RID specular;
+		RID normal_buffer;
+		RID roughness_buffer;
+		RID depth_fb;
+		RID depth_normal_fb;
+		RID depth_normal_roughness_fb;
+		RID color_fb;
+		RID color_specular_fb;
+		int width, height;
+
+		void ensure_specular();
+		void clear();
+		virtual void configure(RID p_color_buffer, RID p_depth_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa);
+
+		RID uniform_set;
+
+		~RenderBufferDataHighEnd();
+	};
+
+	virtual RenderBufferData *_create_render_buffer_data();
+	void _allocate_normal_texture(RenderBufferDataHighEnd *rb);
+	void _allocate_roughness_texture(RenderBufferDataHighEnd *rb);
+
+	RID shadow_sampler;
+	RID render_base_uniform_set;
+	RID view_dependant_uniform_set;
+
+	virtual void _base_uniforms_changed();
+	void _render_buffers_clear_uniform_set(RenderBufferDataHighEnd *rb);
+	virtual void _render_buffers_uniform_set_changed(RID p_render_buffers);
+	virtual RID _render_buffers_get_roughness_texture(RID p_render_buffers);
+	virtual RID _render_buffers_get_normal_texture(RID p_render_buffers);
+
+	void _update_render_base_uniform_set();
+	void _setup_view_dependant_uniform_set(RID p_shadow_atlas, RID p_reflection_atlas);
+	void _update_render_buffers_uniform_set(RID p_render_buffers);
+
+	/* Scene State UBO */
+
+	struct ReflectionData { //should always be 128 bytes
+		float box_extents[3];
+		float index;
+		float box_offset[3];
+		uint32_t mask;
+		float params[4]; // intensity, 0, interior , boxproject
+		float ambient[4]; // ambient color, energy
+		float local_matrix[16]; // up to here for spot and omni, rest is for directional
+	};
+
+	struct LightData {
+		float position[3];
+		float inv_radius;
+		float direction[3];
+		uint16_t attenuation_energy[2]; //16 bits attenuation, then energy
+		uint8_t color_specular[4]; //rgb color, a specular (8 bit unorm)
+		uint16_t cone_attenuation_angle[2]; // attenuation and angle, (16bit float)
+		uint32_t mask;
+		uint8_t shadow_color_enabled[4]; //shadow rgb color, a>0.5 enabled (8bit unorm)
+		float atlas_rect[4]; // in omni, used for atlas uv, in spot, used for projector uv
+		float shadow_matrix[16];
+	};
+
+	struct DirectionalLightData {
+
+		float direction[3];
+		float energy;
+		float color[3];
+		float specular;
+		float shadow_color[3];
+		uint32_t mask;
+		uint32_t blend_splits;
+		uint32_t shadow_enabled;
+		float fade_from;
+		float fade_to;
+		float shadow_split_offsets[4];
+		float shadow_matrices[4][16];
+	};
+
+	struct GIProbeData {
+		float xform[16];
+		float bounds[3];
+		float dynamic_range;
+
+		float bias;
+		float normal_bias;
+		uint32_t blend_ambient;
+		uint32_t texture_slot;
+
+		float anisotropy_strength;
+		float ao;
+		float ao_size;
+		uint32_t pad[1];
+	};
+
+	enum {
+		INSTANCE_DATA_FLAG_MULTIMESH = 1 << 12,
+		INSTANCE_DATA_FLAG_MULTIMESH_FORMAT_2D = 1 << 13,
+		INSTANCE_DATA_FLAG_MULTIMESH_HAS_COLOR = 1 << 14,
+		INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA = 1 << 15,
+		INSTANCE_DATA_FLAGS_MULTIMESH_STRIDE_SHIFT = 16,
+		INSTANCE_DATA_FLAGS_MULTIMESH_STRIDE_MASK = 0x7,
+		INSTANCE_DATA_FLAG_SKELETON = 1 << 19,
+	};
+
+	struct InstanceData {
+		float transform[16];
+		float normal_transform[16];
+		uint32_t flags;
+		uint32_t instance_ofs; //instance_offset in instancing/skeleton buffer
+		uint32_t gi_offset; //GI information when using lightmapping (VCT or lightmap)
+		uint32_t mask;
+	};
+
+	struct SceneState {
+		struct UBO {
+			float projection_matrix[16];
+			float inv_projection_matrix[16];
+
+			float camera_matrix[16];
+			float inv_camera_matrix[16];
+
+			float viewport_size[2];
+			float screen_pixel_size[2];
+
+			float shadow_z_offset;
+			float shadow_z_slope_scale;
+
+			float time;
+			float reflection_multiplier;
+
+			float ambient_light_color_energy[4];
+
+			float ambient_color_sky_mix;
+			uint32_t use_ambient_light;
+			uint32_t use_ambient_cubemap;
+			uint32_t use_reflection_cubemap;
+
+			float radiance_inverse_xform[12];
+
+			float shadow_atlas_pixel_size[2];
+			float directional_shadow_pixel_size[2];
+
+			uint32_t directional_light_count;
+			float dual_paraboloid_side;
+			float z_far;
+			float z_near;
+
+			uint32_t ssao_enabled;
+			float ssao_light_affect;
+			float ssao_ao_affect;
+			uint32_t roughness_limiter_enabled;
+
+			float ao_color[4];
+		};
+
+		UBO ubo;
+
+		RID uniform_buffer;
+
+		ReflectionData *reflections;
+		uint32_t max_reflections;
+		RID reflection_buffer;
+		uint32_t max_reflection_probes_per_instance;
+
+		GIProbeData *gi_probes;
+		uint32_t max_gi_probes;
+		RID gi_probe_buffer;
+		uint32_t max_gi_probe_probes_per_instance;
+
+		LightData *lights;
+		uint32_t max_lights;
+		RID light_buffer;
+
+		DirectionalLightData *directional_lights;
+		uint32_t max_directional_lights;
+		RID directional_light_buffer;
+
+		RID instance_buffer;
+		InstanceData *instances;
+		uint32_t max_instances;
+
+		bool used_screen_texture = false;
+		bool used_normal_texture = false;
+		bool used_depth_texture = false;
+		bool used_sss = false;
+		uint32_t current_shader_index = 0;
+		uint32_t current_material_index = 0;
+	} scene_state;
+
+	/* Render List */
+
+	struct RenderList {
+
+		int max_elements;
+
+		struct Element {
+			RasterizerScene::InstanceBase *instance;
+			MaterialData *material;
+			union {
+				struct {
+					//from least significant to most significant in sort, TODO: should be endian swapped on big endian
+					uint64_t geometry_index : 20;
+					uint64_t material_index : 15;
+					uint64_t shader_index : 12;
+					uint64_t uses_instancing : 1;
+					uint64_t uses_vct : 1;
+					uint64_t uses_lightmap : 1;
+					uint64_t depth_layer : 4;
+					uint64_t priority : 8;
+				};
+
+				uint64_t sort_key;
+			};
+			uint32_t surface_index;
+		};
+
+		Element *base_elements;
+		Element **elements;
+
+		int element_count;
+		int alpha_element_count;
+
+		void clear() {
+
+			element_count = 0;
+			alpha_element_count = 0;
+		}
+
+		//should eventually be replaced by radix
+
+		struct SortByKey {
+
+			_FORCE_INLINE_ bool operator()(const Element *A, const Element *B) const {
+				return A->sort_key < B->sort_key;
+			}
+		};
+
+		void sort_by_key(bool p_alpha) {
+
+			SortArray<Element *, SortByKey> sorter;
+			if (p_alpha) {
+				sorter.sort(&elements[max_elements - alpha_element_count], alpha_element_count);
+			} else {
+				sorter.sort(elements, element_count);
+			}
+		}
+
+		struct SortByDepth {
+
+			_FORCE_INLINE_ bool operator()(const Element *A, const Element *B) const {
+				return A->instance->depth < B->instance->depth;
+			}
+		};
+
+		void sort_by_depth(bool p_alpha) { //used for shadows
+
+			SortArray<Element *, SortByDepth> sorter;
+			if (p_alpha) {
+				sorter.sort(&elements[max_elements - alpha_element_count], alpha_element_count);
+			} else {
+				sorter.sort(elements, element_count);
+			}
+		}
+
+		struct SortByReverseDepthAndPriority {
+
+			_FORCE_INLINE_ bool operator()(const Element *A, const Element *B) const {
+				uint32_t layer_A = uint32_t(A->priority);
+				uint32_t layer_B = uint32_t(B->priority);
+				if (layer_A == layer_B) {
+					return A->instance->depth > B->instance->depth;
+				} else {
+					return layer_A < layer_B;
+				}
+			}
+		};
+
+		void sort_by_reverse_depth_and_priority(bool p_alpha) { //used for alpha
+
+			SortArray<Element *, SortByReverseDepthAndPriority> sorter;
+			if (p_alpha) {
+				sorter.sort(&elements[max_elements - alpha_element_count], alpha_element_count);
+			} else {
+				sorter.sort(elements, element_count);
+			}
+		}
+
+		_FORCE_INLINE_ Element *add_element() {
+
+			if (element_count + alpha_element_count >= max_elements)
+				return NULL;
+			elements[element_count] = &base_elements[element_count];
+			return elements[element_count++];
+		}
+
+		_FORCE_INLINE_ Element *add_alpha_element() {
+
+			if (element_count + alpha_element_count >= max_elements)
+				return NULL;
+			int idx = max_elements - alpha_element_count - 1;
+			elements[idx] = &base_elements[idx];
+			alpha_element_count++;
+			return elements[idx];
+		}
+
+		void init() {
+
+			element_count = 0;
+			alpha_element_count = 0;
+			elements = memnew_arr(Element *, max_elements);
+			base_elements = memnew_arr(Element, max_elements);
+			for (int i = 0; i < max_elements; i++)
+				elements[i] = &base_elements[i]; // assign elements
+		}
+
+		RenderList() {
+
+			max_elements = 0;
+		}
+
+		~RenderList() {
+			memdelete_arr(elements);
+			memdelete_arr(base_elements);
+		}
+	};
+
+	RenderList render_list;
+
+	static RasterizerSceneHighEndRD *singleton;
+	uint64_t render_pass;
+	double time;
+	RID default_shader;
+	RID default_material;
+	RID overdraw_material_shader;
+	RID overdraw_material;
+	RID wireframe_material_shader;
+	RID wireframe_material;
+	RID default_shader_rd;
+	RID default_radiance_uniform_set;
+	RID default_render_buffers_uniform_set;
+
+	RID default_vec4_xform_buffer;
+	RID default_vec4_xform_uniform_set;
+
+	LightClusterBuilder cluster_builder;
+
+	enum PassMode {
+		PASS_MODE_COLOR,
+		PASS_MODE_COLOR_SPECULAR,
+		PASS_MODE_COLOR_TRANSPARENT,
+		PASS_MODE_SHADOW,
+		PASS_MODE_SHADOW_DP,
+		PASS_MODE_DEPTH,
+		PASS_MODE_DEPTH_NORMAL,
+		PASS_MODE_DEPTH_NORMAL_ROUGHNESS,
+		PASS_MODE_DEPTH_MATERIAL,
+	};
+
+	void _setup_environment(RID p_environment, const CameraMatrix &p_cam_projection, const Transform &p_cam_transform, RID p_reflection_probe, bool p_no_fog, const Size2 &p_screen_pixel_size, RID p_shadow_atlas, bool p_flip_y, const Color &p_default_bg_color, float p_znear, float p_zfar, bool p_opaque_render_buffers = false);
+	void _setup_lights(RID *p_light_cull_result, int p_light_cull_count, const Transform &p_camera_inverse_transform, RID p_shadow_atlas, bool p_using_shadows);
+	void _setup_reflections(RID *p_reflection_probe_cull_result, int p_reflection_probe_cull_count, const Transform &p_camera_inverse_transform, RID p_environment);
+	void _setup_gi_probes(RID *p_gi_probe_probe_cull_result, int p_gi_probe_probe_cull_count, const Transform &p_camera_transform);
+
+	void _fill_instances(RenderList::Element **p_elements, int p_element_count, bool p_for_depth);
+	void _render_list(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderList::Element **p_elements, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, bool p_no_gi, RID p_radiance_uniform_set, RID p_render_buffers_uniform_set);
+	_FORCE_INLINE_ void _add_geometry(InstanceBase *p_instance, uint32_t p_surface, RID p_material, PassMode p_pass_mode, uint32_t p_geometry_index);
+	_FORCE_INLINE_ void _add_geometry_with_material(InstanceBase *p_instance, uint32_t p_surface, MaterialData *p_material, RID p_material_rid, PassMode p_pass_mode, uint32_t p_geometry_index);
+
+	void _fill_render_list(InstanceBase **p_cull_result, int p_cull_count, PassMode p_pass_mode, bool p_no_gi);
+
+protected:
+	virtual void _render_scene(RID p_render_buffer, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, RID *p_light_cull_result, int p_light_cull_count, RID *p_reflection_probe_cull_result, int p_reflection_probe_cull_count, RID *p_gi_probe_cull_result, int p_gi_probe_cull_count, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, const Color &p_default_bg_color);
+	virtual void _render_shadow(RID p_framebuffer, InstanceBase **p_cull_result, int p_cull_count, const CameraMatrix &p_projection, const Transform &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip);
+	virtual void _render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, InstanceBase **p_cull_result, int p_cull_count, RID p_framebuffer, const Rect2i &p_region);
+
+public:
+	virtual void set_time(double p_time, double p_step);
+
+	virtual bool free(RID p_rid);
+
+	RasterizerSceneHighEndRD(RasterizerStorageRD *p_storage);
+	~RasterizerSceneHighEndRD();
+};
+#endif // RASTERIZER_SCENE_HIGHEND_RD_H

servers/rendering/rasterizer_rd/render_pipeline_vertex_format_cache_rd.cpp

@@ -0,0 +1,97 @@
+/*************************************************************************/
+/*  render_pipeline_vertex_format_cache_rd.cpp                           */
+/*************************************************************************/
+/*                       This file is part of:                           */
+/*                           GODOT ENGINE                                */
+/*                      https://godotengine.org                          */
+/*************************************************************************/
+/* Copyright (c) 2007-2020 Juan Linietsky, Ariel Manzur.                 */
+/* Copyright (c) 2014-2020 Godot Engine contributors (cf. AUTHORS.md).   */
+/*                                                                       */
+/* Permission is hereby granted, free of charge, to any person obtaining */
+/* a copy of this software and associated documentation files (the       */
+/* "Software"), to deal in the Software without restriction, including   */
+/* without limitation the rights to use, copy, modify, merge, publish,   */
+/* distribute, sublicense, and/or sell copies of the Software, and to    */
+/* permit persons to whom the Software is furnished to do so, subject to */
+/* the following conditions:                                             */
+/*                                                                       */
+/* The above copyright notice and this permission notice shall be        */
+/* included in all copies or substantial portions of the Software.       */
+/*                                                                       */
+/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,       */
+/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF    */
+/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
+/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY  */
+/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,  */
+/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE     */
+/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                */
+/*************************************************************************/
+
+#include "render_pipeline_vertex_format_cache_rd.h"
+#include "core/os/memory.h"
+
+RID RenderPipelineVertexFormatCacheRD::_generate_version(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id) {
+
+	RD::PipelineMultisampleState multisample_state_version = multisample_state;
+	multisample_state_version.sample_count = RD::get_singleton()->framebuffer_format_get_texture_samples(p_framebuffer_format_id);
+
+	RID pipeline = RD::get_singleton()->render_pipeline_create(shader, p_framebuffer_format_id, p_vertex_format_id, render_primitive, rasterization_state, multisample_state_version, depth_stencil_state, blend_state, dynamic_state_flags);
+	ERR_FAIL_COND_V(pipeline.is_null(), RID());
+	versions = (Version *)memrealloc(versions, sizeof(Version) * (version_count + 1));
+	versions[version_count].framebuffer_id = p_framebuffer_format_id;
+	versions[version_count].vertex_id = p_vertex_format_id;
+	versions[version_count].pipeline = pipeline;
+	version_count++;
+	return pipeline;
+}
+
+void RenderPipelineVertexFormatCacheRD::_clear() {
+
+	if (versions) {
+		for (uint32_t i = 0; i < version_count; i++) {
+			//shader may be gone, so this may not be valid
+			if (RD::get_singleton()->render_pipeline_is_valid(versions[i].pipeline)) {
+				RD::get_singleton()->free(versions[i].pipeline);
+			}
+		}
+		version_count = 0;
+		memfree(versions);
+		versions = NULL;
+	}
+}
+
+void RenderPipelineVertexFormatCacheRD::setup(RID p_shader, RD::RenderPrimitive p_primitive, const RD::PipelineRasterizationState &p_rasterization_state, RD::PipelineMultisampleState p_multisample, const RD::PipelineDepthStencilState &p_depth_stencil_state, const RD::PipelineColorBlendState &p_blend_state, int p_dynamic_state_flags) {
+	ERR_FAIL_COND(p_shader.is_null());
+	_clear();
+	shader = p_shader;
+	input_mask = RD::get_singleton()->shader_get_vertex_input_attribute_mask(p_shader);
+	render_primitive = p_primitive;
+	rasterization_state = p_rasterization_state;
+	multisample_state = p_multisample;
+	depth_stencil_state = p_depth_stencil_state;
+	blend_state = p_blend_state;
+	dynamic_state_flags = p_dynamic_state_flags;
+}
+
+void RenderPipelineVertexFormatCacheRD::update_shader(RID p_shader) {
+	ERR_FAIL_COND(p_shader.is_null());
+	_clear();
+	setup(p_shader, render_primitive, rasterization_state, multisample_state, depth_stencil_state, blend_state, dynamic_state_flags);
+}
+
+void RenderPipelineVertexFormatCacheRD::clear() {
+	_clear();
+	shader = RID(); //clear shader
+	input_mask = 0;
+}
+
+RenderPipelineVertexFormatCacheRD::RenderPipelineVertexFormatCacheRD() {
+	version_count = 0;
+	versions = NULL;
+	input_mask = 0;
+}
+
+RenderPipelineVertexFormatCacheRD::~RenderPipelineVertexFormatCacheRD() {
+	_clear();
+}

servers/rendering/rasterizer_rd/render_pipeline_vertex_format_cache_rd.h

@@ -0,0 +1,96 @@
+/*************************************************************************/
+/*  render_pipeline_vertex_format_cache_rd.h                             */
+/*************************************************************************/
+/*                       This file is part of:                           */
+/*                           GODOT ENGINE                                */
+/*                      https://godotengine.org                          */
+/*************************************************************************/
+/* Copyright (c) 2007-2020 Juan Linietsky, Ariel Manzur.                 */
+/* Copyright (c) 2014-2020 Godot Engine contributors (cf. AUTHORS.md).   */
+/*                                                                       */
+/* Permission is hereby granted, free of charge, to any person obtaining */
+/* a copy of this software and associated documentation files (the       */
+/* "Software"), to deal in the Software without restriction, including   */
+/* without limitation the rights to use, copy, modify, merge, publish,   */
+/* distribute, sublicense, and/or sell copies of the Software, and to    */
+/* permit persons to whom the Software is furnished to do so, subject to */
+/* the following conditions:                                             */
+/*                                                                       */
+/* The above copyright notice and this permission notice shall be        */
+/* included in all copies or substantial portions of the Software.       */
+/*                                                                       */
+/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,       */
+/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF    */
+/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
+/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY  */
+/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,  */
+/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE     */
+/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                */
+/*************************************************************************/
+
+#ifndef RENDER_PIPELINE_CACHE_RD_H
+#define RENDER_PIPELINE_CACHE_RD_H
+
+#include "core/spin_lock.h"
+#include "servers/rendering/rendering_device.h"
+
+class RenderPipelineVertexFormatCacheRD {
+
+	SpinLock spin_lock;
+
+	RID shader;
+	uint32_t input_mask;
+
+	RD::RenderPrimitive render_primitive;
+	RD::PipelineRasterizationState rasterization_state;
+	RD::PipelineMultisampleState multisample_state;
+	RD::PipelineDepthStencilState depth_stencil_state;
+	RD::PipelineColorBlendState blend_state;
+	int dynamic_state_flags;
+
+	struct Version {
+		RD::VertexFormatID vertex_id;
+		RD::FramebufferFormatID framebuffer_id;
+		RID pipeline;
+	};
+
+	Version *versions;
+	uint32_t version_count;
+
+	RID _generate_version(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id);
+
+	void _clear();
+
+public:
+	void setup(RID p_shader, RD::RenderPrimitive p_primitive, const RD::PipelineRasterizationState &p_rasterization_state, RD::PipelineMultisampleState p_multisample, const RD::PipelineDepthStencilState &p_depth_stencil_state, const RD::PipelineColorBlendState &p_blend_state, int p_dynamic_state_flags = 0);
+	void update_shader(RID p_shader);
+
+	_FORCE_INLINE_ RID get_render_pipeline(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id) {
+#ifdef DEBUG_ENABLED
+		ERR_FAIL_COND_V_MSG(shader.is_null(), RID(),
+				"Attempted to use an unused shader variant (shader is null),");
+#endif
+
+		spin_lock.lock();
+		RID result;
+		for (uint32_t i = 0; i < version_count; i++) {
+			if (versions[i].vertex_id == p_vertex_format_id && versions[i].framebuffer_id == p_framebuffer_format_id) {
+				result = versions[i].pipeline;
+				spin_lock.unlock();
+				return result;
+			}
+		}
+		result = _generate_version(p_vertex_format_id, p_framebuffer_format_id);
+		spin_lock.unlock();
+		return result;
+	}
+
+	_FORCE_INLINE_ uint32_t get_vertex_input_mask() const {
+		return input_mask;
+	}
+	void clear();
+	RenderPipelineVertexFormatCacheRD();
+	~RenderPipelineVertexFormatCacheRD();
+};
+
+#endif // RENDER_PIPELINE_CACHE_RD_H

servers/rendering/rasterizer_rd/shader_compiler_rd.h

@@ -0,0 +1,123 @@
+/*************************************************************************/
+/*  shader_compiler_rd.h                                                 */
+/*************************************************************************/
+/*                       This file is part of:                           */
+/*                           GODOT ENGINE                                */
+/*                      https://godotengine.org                          */
+/*************************************************************************/
+/* Copyright (c) 2007-2020 Juan Linietsky, Ariel Manzur.                 */
+/* Copyright (c) 2014-2020 Godot Engine contributors (cf. AUTHORS.md).   */
+/*                                                                       */
+/* Permission is hereby granted, free of charge, to any person obtaining */
+/* a copy of this software and associated documentation files (the       */
+/* "Software"), to deal in the Software without restriction, including   */
+/* without limitation the rights to use, copy, modify, merge, publish,   */
+/* distribute, sublicense, and/or sell copies of the Software, and to    */
+/* permit persons to whom the Software is furnished to do so, subject to */
+/* the following conditions:                                             */
+/*                                                                       */
+/* The above copyright notice and this permission notice shall be        */
+/* included in all copies or substantial portions of the Software.       */
+/*                                                                       */
+/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,       */
+/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF    */
+/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
+/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY  */
+/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,  */
+/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE     */
+/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                */
+/*************************************************************************/
+
+#ifndef SHADER_COMPILER_RD_H
+#define SHADER_COMPILER_RD_H
+
+#include "core/pair.h"
+#include "servers/rendering/shader_language.h"
+#include "servers/rendering/shader_types.h"
+#include "servers/rendering_server.h"
+
+class ShaderCompilerRD {
+public:
+	struct IdentifierActions {
+
+		Map<StringName, Pair<int *, int>> render_mode_values;
+		Map<StringName, bool *> render_mode_flags;
+		Map<StringName, bool *> usage_flag_pointers;
+		Map<StringName, bool *> write_flag_pointers;
+
+		Map<StringName, ShaderLanguage::ShaderNode::Uniform> *uniforms;
+	};
+
+	struct GeneratedCode {
+
+		Vector<String> defines;
+		struct Texture {
+			StringName name;
+			ShaderLanguage::DataType type;
+			ShaderLanguage::ShaderNode::Uniform::Hint hint;
+			ShaderLanguage::TextureFilter filter;
+			ShaderLanguage::TextureRepeat repeat;
+		};
+
+		Vector<Texture> texture_uniforms;
+
+		Vector<uint32_t> uniform_offsets;
+		uint32_t uniform_total_size;
+		String uniforms;
+		String vertex_global;
+		String vertex;
+		String fragment_global;
+		String fragment;
+		String light;
+
+		bool uses_fragment_time;
+		bool uses_vertex_time;
+	};
+
+	struct DefaultIdentifierActions {
+
+		Map<StringName, String> renames;
+		Map<StringName, String> render_mode_defines;
+		Map<StringName, String> usage_defines;
+		Map<StringName, String> custom_samplers;
+		ShaderLanguage::TextureFilter default_filter;
+		ShaderLanguage::TextureRepeat default_repeat;
+		String sampler_array_name;
+		int base_texture_binding_index = 0;
+		int texture_layout_set = 0;
+		String base_uniform_string;
+		uint32_t base_varying_index = 0;
+	};
+
+private:
+	ShaderLanguage parser;
+
+	String _get_sampler_name(ShaderLanguage::TextureFilter p_filter, ShaderLanguage::TextureRepeat p_repeat);
+
+	void _dump_function_deps(const ShaderLanguage::ShaderNode *p_node, const StringName &p_for_func, const Map<StringName, String> &p_func_code, String &r_to_add, Set<StringName> &added);
+	String _dump_node_code(const ShaderLanguage::Node *p_node, int p_level, GeneratedCode &r_gen_code, IdentifierActions &p_actions, const DefaultIdentifierActions &p_default_actions, bool p_assigning);
+
+	const ShaderLanguage::ShaderNode *shader;
+	const ShaderLanguage::FunctionNode *function;
+	StringName current_func_name;
+	StringName vertex_name;
+	StringName fragment_name;
+	StringName light_name;
+	StringName time_name;
+	Set<StringName> texture_functions;
+
+	Set<StringName> used_name_defines;
+	Set<StringName> used_flag_pointers;
+	Set<StringName> used_rmode_defines;
+	Set<StringName> internal_functions;
+
+	DefaultIdentifierActions actions;
+
+public:
+	Error compile(RS::ShaderMode p_mode, const String &p_code, IdentifierActions *p_actions, const String &p_path, GeneratedCode &r_gen_code);
+
+	void initialize(DefaultIdentifierActions p_actions);
+	ShaderCompilerRD();
+};
+
+#endif // SHADERCOMPILERRD_H

servers/rendering/rasterizer_rd/shader_rd.cpp

@@ -0,0 +1,495 @@
+/*************************************************************************/
+/*  shader_rd.cpp                                                        */
+/*************************************************************************/
+/*                       This file is part of:                           */
+/*                           GODOT ENGINE                                */
+/*                      https://godotengine.org                          */
+/*************************************************************************/
+/* Copyright (c) 2007-2020 Juan Linietsky, Ariel Manzur.                 */
+/* Copyright (c) 2014-2020 Godot Engine contributors (cf. AUTHORS.md).   */
+/*                                                                       */
+/* Permission is hereby granted, free of charge, to any person obtaining */
+/* a copy of this software and associated documentation files (the       */
+/* "Software"), to deal in the Software without restriction, including   */
+/* without limitation the rights to use, copy, modify, merge, publish,   */
+/* distribute, sublicense, and/or sell copies of the Software, and to    */
+/* permit persons to whom the Software is furnished to do so, subject to */
+/* the following conditions:                                             */
+/*                                                                       */
+/* The above copyright notice and this permission notice shall be        */
+/* included in all copies or substantial portions of the Software.       */
+/*                                                                       */
+/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,       */
+/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF    */
+/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
+/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY  */
+/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,  */
+/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE     */
+/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                */
+/*************************************************************************/
+
+#include "shader_rd.h"
+
+#include "core/string_builder.h"
+#include "rasterizer_rd.h"
+#include "servers/rendering/rendering_device.h"
+
+void ShaderRD::setup(const char *p_vertex_code, const char *p_fragment_code, const char *p_compute_code, const char *p_name) {
+
+	name = p_name;
+	//split vertex and shader code (thank you, shader compiler programmers from you know what company).
+	if (p_vertex_code) {
+		String defines_tag = "\nVERSION_DEFINES";
+		String globals_tag = "\nVERTEX_SHADER_GLOBALS";
+		String material_tag = "\nMATERIAL_UNIFORMS";
+		String code_tag = "\nVERTEX_SHADER_CODE";
+		String code = p_vertex_code;
+
+		int cpos = code.find(defines_tag);
+		if (cpos != -1) {
+			vertex_codev = code.substr(0, cpos).ascii();
+			code = code.substr(cpos + defines_tag.length(), code.length());
+		}
+
+		cpos = code.find(material_tag);
+
+		if (cpos == -1) {
+			vertex_code0 = code.ascii();
+		} else {
+			vertex_code0 = code.substr(0, cpos).ascii();
+			code = code.substr(cpos + material_tag.length(), code.length());
+
+			cpos = code.find(globals_tag);
+
+			if (cpos == -1) {
+				vertex_code1 = code.ascii();
+			} else {
+
+				vertex_code1 = code.substr(0, cpos).ascii();
+				String code2 = code.substr(cpos + globals_tag.length(), code.length());
+
+				cpos = code2.find(code_tag);
+				if (cpos == -1) {
+					vertex_code2 = code2.ascii();
+				} else {
+
+					vertex_code2 = code2.substr(0, cpos).ascii();
+					vertex_code3 = code2.substr(cpos + code_tag.length(), code2.length()).ascii();
+				}
+			}
+		}
+	}
+
+	if (p_fragment_code) {
+		String defines_tag = "\nVERSION_DEFINES";
+		String globals_tag = "\nFRAGMENT_SHADER_GLOBALS";
+		String material_tag = "\nMATERIAL_UNIFORMS";
+		String code_tag = "\nFRAGMENT_SHADER_CODE";
+		String light_code_tag = "\nLIGHT_SHADER_CODE";
+		String code = p_fragment_code;
+
+		int cpos = code.find(defines_tag);
+		if (cpos != -1) {
+			fragment_codev = code.substr(0, cpos).ascii();
+			code = code.substr(cpos + defines_tag.length(), code.length());
+		}
+
+		cpos = code.find(material_tag);
+		if (cpos == -1) {
+			fragment_code0 = code.ascii();
+		} else {
+			fragment_code0 = code.substr(0, cpos).ascii();
+			//print_line("CODE0:\n"+String(fragment_code0.get_data()));
+			code = code.substr(cpos + material_tag.length(), code.length());
+			cpos = code.find(globals_tag);
+
+			if (cpos == -1) {
+				fragment_code1 = code.ascii();
+			} else {
+
+				fragment_code1 = code.substr(0, cpos).ascii();
+				//print_line("CODE1:\n"+String(fragment_code1.get_data()));
+
+				String code2 = code.substr(cpos + globals_tag.length(), code.length());
+				cpos = code2.find(light_code_tag);
+
+				if (cpos == -1) {
+					fragment_code2 = code2.ascii();
+				} else {
+
+					fragment_code2 = code2.substr(0, cpos).ascii();
+					//print_line("CODE2:\n"+String(fragment_code2.get_data()));
+
+					String code3 = code2.substr(cpos + light_code_tag.length(), code2.length());
+
+					cpos = code3.find(code_tag);
+					if (cpos == -1) {
+						fragment_code3 = code3.ascii();
+					} else {
+
+						fragment_code3 = code3.substr(0, cpos).ascii();
+						//print_line("CODE3:\n"+String(fragment_code3.get_data()));
+						fragment_code4 = code3.substr(cpos + code_tag.length(), code3.length()).ascii();
+						//print_line("CODE4:\n"+String(fragment_code4.get_data()));
+					}
+				}
+			}
+		}
+	}
+
+	if (p_compute_code) {
+		is_compute = true;
+
+		String defines_tag = "\nVERSION_DEFINES";
+		String globals_tag = "\nCOMPUTE_SHADER_GLOBALS";
+		String material_tag = "\nMATERIAL_UNIFORMS";
+		String code_tag = "\nCOMPUTE_SHADER_CODE";
+		String code = p_compute_code;
+
+		int cpos = code.find(defines_tag);
+		if (cpos != -1) {
+			compute_codev = code.substr(0, cpos).ascii();
+			code = code.substr(cpos + defines_tag.length(), code.length());
+		}
+
+		cpos = code.find(material_tag);
+
+		if (cpos == -1) {
+			compute_code0 = code.ascii();
+		} else {
+			compute_code0 = code.substr(0, cpos).ascii();
+			code = code.substr(cpos + material_tag.length(), code.length());
+
+			cpos = code.find(globals_tag);
+
+			if (cpos == -1) {
+				compute_code1 = code.ascii();
+			} else {
+
+				compute_code1 = code.substr(0, cpos).ascii();
+				String code2 = code.substr(cpos + globals_tag.length(), code.length());
+
+				cpos = code2.find(code_tag);
+				if (cpos == -1) {
+					compute_code2 = code2.ascii();
+				} else {
+
+					compute_code2 = code2.substr(0, cpos).ascii();
+					compute_code3 = code2.substr(cpos + code_tag.length(), code2.length()).ascii();
+				}
+			}
+		}
+	}
+}
+
+RID ShaderRD::version_create() {
+
+	//initialize() was never called
+	ERR_FAIL_COND_V(variant_defines.size() == 0, RID());
+
+	Version version;
+	version.dirty = true;
+	version.valid = false;
+	version.initialize_needed = true;
+	version.variants = NULL;
+	return version_owner.make_rid(version);
+}
+
+void ShaderRD::_clear_version(Version *p_version) {
+	//clear versions if they exist
+	if (p_version->variants) {
+		for (int i = 0; i < variant_defines.size(); i++) {
+			RD::get_singleton()->free(p_version->variants[i]);
+		}
+
+		memdelete_arr(p_version->variants);
+		p_version->variants = NULL;
+	}
+}
+
+void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) {
+
+	Vector<RD::ShaderStageData> stages;
+
+	String error;
+	String current_source;
+	RD::ShaderStage current_stage = RD::SHADER_STAGE_VERTEX;
+	bool build_ok = true;
+
+	if (!is_compute) {
+		//vertex stage
+
+		StringBuilder builder;
+
+		builder.append(vertex_codev.get_data()); // version info (if exists)
+		builder.append("\n"); //make sure defines begin at newline
+		builder.append(general_defines.get_data());
+		builder.append(variant_defines[p_variant].get_data());
+
+		for (int j = 0; j < p_version->custom_defines.size(); j++) {
+			builder.append(p_version->custom_defines[j].get_data());
+		}
+
+		builder.append(vertex_code0.get_data()); //first part of vertex
+
+		builder.append(p_version->uniforms.get_data()); //uniforms (same for vertex and fragment)
+
+		builder.append(vertex_code1.get_data()); //second part of vertex
+
+		builder.append(p_version->vertex_globals.get_data()); // vertex globals
+
+		builder.append(vertex_code2.get_data()); //third part of vertex
+
+		builder.append(p_version->vertex_code.get_data()); // code
+
+		builder.append(vertex_code3.get_data()); //fourth of vertex
+
+		current_source = builder.as_string();
+		RD::ShaderStageData stage;
+		stage.spir_v = RD::get_singleton()->shader_compile_from_source(RD::SHADER_STAGE_VERTEX, current_source, RD::SHADER_LANGUAGE_GLSL, &error);
+		if (stage.spir_v.size() == 0) {
+			build_ok = false;
+		} else {
+
+			stage.shader_stage = RD::SHADER_STAGE_VERTEX;
+			stages.push_back(stage);
+		}
+	}
+
+	if (!is_compute && build_ok) {
+		//fragment stage
+		current_stage = RD::SHADER_STAGE_FRAGMENT;
+
+		StringBuilder builder;
+
+		builder.append(fragment_codev.get_data()); // version info (if exists)
+		builder.append("\n"); //make sure defines begin at newline
+
+		builder.append(general_defines.get_data());
+		builder.append(variant_defines[p_variant].get_data());
+		for (int j = 0; j < p_version->custom_defines.size(); j++) {
+			builder.append(p_version->custom_defines[j].get_data());
+		}
+
+		builder.append(fragment_code0.get_data()); //first part of fragment
+
+		builder.append(p_version->uniforms.get_data()); //uniforms (same for fragment and fragment)
+
+		builder.append(fragment_code1.get_data()); //first part of fragment
+
+		builder.append(p_version->fragment_globals.get_data()); // fragment globals
+
+		builder.append(fragment_code2.get_data()); //third part of fragment
+
+		builder.append(p_version->fragment_light.get_data()); // fragment light
+
+		builder.append(fragment_code3.get_data()); //fourth part of fragment
+
+		builder.append(p_version->fragment_code.get_data()); // fragment code
+
+		builder.append(fragment_code4.get_data()); //fourth part of fragment
+
+		current_source = builder.as_string();
+		RD::ShaderStageData stage;
+		stage.spir_v = RD::get_singleton()->shader_compile_from_source(RD::SHADER_STAGE_FRAGMENT, current_source, RD::SHADER_LANGUAGE_GLSL, &error);
+		if (stage.spir_v.size() == 0) {
+			build_ok = false;
+		} else {
+
+			stage.shader_stage = RD::SHADER_STAGE_FRAGMENT;
+			stages.push_back(stage);
+		}
+	}
+
+	if (is_compute) {
+		//compute stage
+		current_stage = RD::SHADER_STAGE_COMPUTE;
+
+		StringBuilder builder;
+
+		builder.append(compute_codev.get_data()); // version info (if exists)
+		builder.append("\n"); //make sure defines begin at newline
+		builder.append(general_defines.get_data());
+		builder.append(variant_defines[p_variant].get_data());
+
+		for (int j = 0; j < p_version->custom_defines.size(); j++) {
+			builder.append(p_version->custom_defines[j].get_data());
+		}
+
+		builder.append(compute_code0.get_data()); //first part of compute
+
+		builder.append(p_version->uniforms.get_data()); //uniforms (same for compute and fragment)
+
+		builder.append(compute_code1.get_data()); //second part of compute
+
+		builder.append(p_version->compute_globals.get_data()); // compute globals
+
+		builder.append(compute_code2.get_data()); //third part of compute
+
+		builder.append(p_version->compute_code.get_data()); // code
+
+		builder.append(compute_code3.get_data()); //fourth of compute
+
+		current_source = builder.as_string();
+		RD::ShaderStageData stage;
+		stage.spir_v = RD::get_singleton()->shader_compile_from_source(RD::SHADER_STAGE_COMPUTE, current_source, RD::SHADER_LANGUAGE_GLSL, &error);
+		if (stage.spir_v.size() == 0) {
+			build_ok = false;
+		} else {
+
+			stage.shader_stage = RD::SHADER_STAGE_COMPUTE;
+			stages.push_back(stage);
+		}
+	}
+
+	if (!build_ok) {
+		MutexLock lock(variant_set_mutex); //properly print the errors
+		ERR_PRINT("Error compiling " + String(current_stage == RD::SHADER_STAGE_COMPUTE ? "Compute " : (current_stage == RD::SHADER_STAGE_VERTEX ? "Vertex" : "Fragment")) + " shader, variant #" + itos(p_variant) + " (" + variant_defines[p_variant].get_data() + ").");
+		ERR_PRINT(error);
+
+#ifdef DEBUG_ENABLED
+		ERR_PRINT("code:\n" + current_source.get_with_code_lines());
+#endif
+		return;
+	}
+
+	RID shader = RD::get_singleton()->shader_create(stages);
+	{
+		MutexLock lock(variant_set_mutex);
+		p_version->variants[p_variant] = shader;
+	}
+}
+
+void ShaderRD::_compile_version(Version *p_version) {
+
+	_clear_version(p_version);
+
+	p_version->valid = false;
+	p_version->dirty = false;
+
+	p_version->variants = memnew_arr(RID, variant_defines.size());
+#if 1
+
+	RasterizerRD::thread_work_pool.do_work(variant_defines.size(), this, &ShaderRD::_compile_variant, p_version);
+#else
+	for (int i = 0; i < variant_defines.size(); i++) {
+
+		_compile_variant(i, p_version);
+	}
+#endif
+
+	bool all_valid = true;
+	for (int i = 0; i < variant_defines.size(); i++) {
+		if (p_version->variants[i].is_null()) {
+			all_valid = false;
+			break;
+		}
+	}
+
+	if (!all_valid) {
+		//clear versions if they exist
+		for (int i = 0; i < variant_defines.size(); i++) {
+			if (!p_version->variants[i].is_null()) {
+				RD::get_singleton()->free(p_version->variants[i]);
+			}
+		}
+		memdelete_arr(p_version->variants);
+		p_version->variants = NULL;
+		return;
+	}
+
+	p_version->valid = true;
+}
+
+void ShaderRD::version_set_code(RID p_version, const String &p_uniforms, const String &p_vertex_globals, const String &p_vertex_code, const String &p_fragment_globals, const String &p_fragment_light, const String &p_fragment_code, const Vector<String> &p_custom_defines) {
+
+	ERR_FAIL_COND(is_compute);
+
+	Version *version = version_owner.getornull(p_version);
+	ERR_FAIL_COND(!version);
+	version->vertex_globals = p_vertex_globals.utf8();
+	version->vertex_code = p_vertex_code.utf8();
+	version->fragment_light = p_fragment_light.utf8();
+	version->fragment_globals = p_fragment_globals.utf8();
+	version->fragment_code = p_fragment_code.utf8();
+	version->uniforms = p_uniforms.utf8();
+
+	version->custom_defines.clear();
+	for (int i = 0; i < p_custom_defines.size(); i++) {
+		version->custom_defines.push_back(p_custom_defines[i].utf8());
+	}
+
+	version->dirty = true;
+	if (version->initialize_needed) {
+		_compile_version(version);
+		version->initialize_needed = false;
+	}
+}
+
+void ShaderRD::version_set_compute_code(RID p_version, const String &p_uniforms, const String &p_compute_globals, const String &p_compute_code, const Vector<String> &p_custom_defines) {
+
+	ERR_FAIL_COND(!is_compute);
+
+	Version *version = version_owner.getornull(p_version);
+	ERR_FAIL_COND(!version);
+	version->compute_globals = p_compute_globals.utf8();
+	version->compute_code = p_compute_code.utf8();
+	version->uniforms = p_uniforms.utf8();
+
+	version->custom_defines.clear();
+	for (int i = 0; i < p_custom_defines.size(); i++) {
+		version->custom_defines.push_back(p_custom_defines[i].utf8());
+	}
+
+	version->dirty = true;
+	if (version->initialize_needed) {
+		_compile_version(version);
+		version->initialize_needed = false;
+	}
+}
+
+bool ShaderRD::version_is_valid(RID p_version) {
+	Version *version = version_owner.getornull(p_version);
+	ERR_FAIL_COND_V(!version, false);
+
+	if (version->dirty) {
+		_compile_version(version);
+	}
+
+	return version->valid;
+}
+
+bool ShaderRD::version_free(RID p_version) {
+
+	if (version_owner.owns(p_version)) {
+		Version *version = version_owner.getornull(p_version);
+		_clear_version(version);
+		version_owner.free(p_version);
+	} else {
+		return false;
+	}
+
+	return true;
+}
+
+void ShaderRD::initialize(const Vector<String> &p_variant_defines, const String &p_general_defines) {
+	ERR_FAIL_COND(variant_defines.size());
+	ERR_FAIL_COND(p_variant_defines.size() == 0);
+	general_defines = p_general_defines.utf8();
+	for (int i = 0; i < p_variant_defines.size(); i++) {
+
+		variant_defines.push_back(p_variant_defines[i].utf8());
+	}
+}
+
+ShaderRD::~ShaderRD() {
+	List<RID> remaining;
+	version_owner.get_owned_list(&remaining);
+	if (remaining.size()) {
+		ERR_PRINT(itos(remaining.size()) + " shaders of type " + name + " were never freed");
+		while (remaining.size()) {
+			version_free(remaining.front()->get());
+			remaining.pop_front();
+		}
+	}
+}

servers/rendering/rasterizer_rd/shader_rd.h

@@ -0,0 +1,136 @@
+/*************************************************************************/
+/*  shader_rd.h                                                          */
+/*************************************************************************/
+/*                       This file is part of:                           */
+/*                           GODOT ENGINE                                */
+/*                      https://godotengine.org                          */
+/*************************************************************************/
+/* Copyright (c) 2007-2020 Juan Linietsky, Ariel Manzur.                 */
+/* Copyright (c) 2014-2020 Godot Engine contributors (cf. AUTHORS.md).   */
+/*                                                                       */
+/* Permission is hereby granted, free of charge, to any person obtaining */
+/* a copy of this software and associated documentation files (the       */
+/* "Software"), to deal in the Software without restriction, including   */
+/* without limitation the rights to use, copy, modify, merge, publish,   */
+/* distribute, sublicense, and/or sell copies of the Software, and to    */
+/* permit persons to whom the Software is furnished to do so, subject to */
+/* the following conditions:                                             */
+/*                                                                       */
+/* The above copyright notice and this permission notice shall be        */
+/* included in all copies or substantial portions of the Software.       */
+/*                                                                       */
+/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,       */
+/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF    */
+/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
+/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY  */
+/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,  */
+/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE     */
+/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                */
+/*************************************************************************/
+
+#ifndef SHADER_RD_H
+#define SHADER_RD_H
+
+#include "core/hash_map.h"
+#include "core/map.h"
+#include "core/os/mutex.h"
+#include "core/rid_owner.h"
+#include "core/variant.h"
+
+#include <stdio.h>
+/**
+	@author Juan Linietsky <reduzio@gmail.com>
+*/
+
+class ShaderRD {
+
+	//versions
+	CharString general_defines;
+	Vector<CharString> variant_defines;
+
+	struct Version {
+		CharString uniforms;
+		CharString vertex_globals;
+		CharString vertex_code;
+		CharString compute_globals;
+		CharString compute_code;
+		CharString fragment_light;
+		CharString fragment_globals;
+		CharString fragment_code;
+		Vector<CharString> custom_defines;
+
+		RID *variants; //same size as version defines
+
+		bool valid;
+		bool dirty;
+		bool initialize_needed;
+	};
+
+	Mutex variant_set_mutex;
+
+	void _compile_variant(uint32_t p_variant, Version *p_version);
+
+	void _clear_version(Version *p_version);
+	void _compile_version(Version *p_version);
+
+	RID_Owner<Version> version_owner;
+
+	CharString fragment_codev; //for version and extensions
+	CharString fragment_code0;
+	CharString fragment_code1;
+	CharString fragment_code2;
+	CharString fragment_code3;
+	CharString fragment_code4;
+
+	CharString vertex_codev; //for version and extensions
+	CharString vertex_code0;
+	CharString vertex_code1;
+	CharString vertex_code2;
+	CharString vertex_code3;
+
+	bool is_compute = false;
+
+	CharString compute_codev; //for version and extensions
+	CharString compute_code0;
+	CharString compute_code1;
+	CharString compute_code2;
+	CharString compute_code3;
+
+	const char *name;
+
+protected:
+	ShaderRD() {}
+	void setup(const char *p_vertex_code, const char *p_fragment_code, const char *p_compute_code, const char *p_name);
+
+public:
+	RID version_create();
+
+	void version_set_code(RID p_version, const String &p_uniforms, const String &p_vertex_globals, const String &p_vertex_code, const String &p_fragment_globals, const String &p_fragment_light, const String &p_fragment_code, const Vector<String> &p_custom_defines);
+	void version_set_compute_code(RID p_version, const String &p_uniforms, const String &p_compute_globals, const String &p_compute_code, const Vector<String> &p_custom_defines);
+
+	_FORCE_INLINE_ RID version_get_shader(RID p_version, int p_variant) {
+		ERR_FAIL_INDEX_V(p_variant, variant_defines.size(), RID());
+
+		Version *version = version_owner.getornull(p_version);
+		ERR_FAIL_COND_V(!version, RID());
+
+		if (version->dirty) {
+			_compile_version(version);
+		}
+
+		if (!version->valid) {
+			return RID();
+		}
+
+		return version->variants[p_variant];
+	}
+
+	bool version_is_valid(RID p_version);
+
+	bool version_free(RID p_version);
+
+	void initialize(const Vector<String> &p_variant_defines, const String &p_general_defines = "");
+	virtual ~ShaderRD();
+};
+
+#endif

servers/rendering/rasterizer_rd/shaders/SCsub

@@ -0,0 +1,24 @@
+#!/usr/bin/env python
+
+Import('env')
+
+if 'RD_GLSL' in env['BUILDERS']:
+    env.RD_GLSL('canvas.glsl');
+    env.RD_GLSL('canvas_occlusion.glsl');
+    env.RD_GLSL('blur.glsl');
+    env.RD_GLSL('cubemap_roughness.glsl');
+    env.RD_GLSL('cubemap_downsampler.glsl');
+    env.RD_GLSL('cubemap_filter.glsl');
+    env.RD_GLSL('scene_high_end.glsl');
+    env.RD_GLSL('sky.glsl');
+    env.RD_GLSL('tonemap.glsl');
+    env.RD_GLSL('copy.glsl');
+    env.RD_GLSL('giprobe.glsl');
+    env.RD_GLSL('giprobe_debug.glsl');
+    env.RD_GLSL('giprobe_sdf.glsl');
+    env.RD_GLSL('luminance_reduce.glsl');
+    env.RD_GLSL('bokeh_dof.glsl');
+    env.RD_GLSL('ssao.glsl');
+    env.RD_GLSL('ssao_minify.glsl');
+    env.RD_GLSL('ssao_blur.glsl');
+    env.RD_GLSL('roughness_limiter.glsl');

servers/rendering/rasterizer_rd/shaders/blur.glsl

@@ -0,0 +1,294 @@
+/* clang-format off */
+[vertex]
+
+#version 450
+
+VERSION_DEFINES
+
+#include "blur_inc.glsl"
+
+layout(location = 0) out vec2 uv_interp;
+/* clang-format on */
+
+void main() {
+
+	vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0));
+	uv_interp = base_arr[gl_VertexIndex];
+
+	if (bool(blur.flags & FLAG_USE_BLUR_SECTION)) {
+		uv_interp = blur.section.xy + uv_interp * blur.section.zw;
+	}
+
+	gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0);
+
+	if (bool(blur.flags & FLAG_FLIP_Y)) {
+		uv_interp.y = 1.0 - uv_interp.y;
+	}
+}
+
+/* clang-format off */
+[fragment]
+
+#version 450
+
+VERSION_DEFINES
+
+#include "blur_inc.glsl"
+
+layout(location = 0) in vec2 uv_interp;
+/* clang-format on */
+
+layout(set = 0, binding = 0) uniform sampler2D source_color;
+
+#ifdef MODE_SSAO_MERGE
+layout(set = 1, binding = 0) uniform sampler2D source_ssao;
+#endif
+
+#ifdef GLOW_USE_AUTO_EXPOSURE
+layout(set = 1, binding = 0) uniform sampler2D source_auto_exposure;
+#endif
+
+layout(location = 0) out vec4 frag_color;
+
+//DOF
+#if defined(MODE_DOF_FAR_BLUR) || defined(MODE_DOF_NEAR_BLUR)
+
+layout(set = 1, binding = 0) uniform sampler2D dof_source_depth;
+
+#ifdef DOF_NEAR_BLUR_MERGE
+layout(set = 2, binding = 0) uniform sampler2D source_dof_original;
+#endif
+
+#ifdef DOF_QUALITY_LOW
+const int dof_kernel_size = 5;
+const int dof_kernel_from = 2;
+const float dof_kernel[5] = float[](0.153388, 0.221461, 0.250301, 0.221461, 0.153388);
+#endif
+
+#ifdef DOF_QUALITY_MEDIUM
+const int dof_kernel_size = 11;
+const int dof_kernel_from = 5;
+const float dof_kernel[11] = float[](0.055037, 0.072806, 0.090506, 0.105726, 0.116061, 0.119726, 0.116061, 0.105726, 0.090506, 0.072806, 0.055037);
+
+#endif
+
+#ifdef DOF_QUALITY_HIGH
+const int dof_kernel_size = 21;
+const int dof_kernel_from = 10;
+const float dof_kernel[21] = float[](0.028174, 0.032676, 0.037311, 0.041944, 0.046421, 0.050582, 0.054261, 0.057307, 0.059587, 0.060998, 0.061476, 0.060998, 0.059587, 0.057307, 0.054261, 0.050582, 0.046421, 0.041944, 0.037311, 0.032676, 0.028174);
+#endif
+
+#endif
+
+void main() {
+
+#ifdef MODE_MIPMAP
+
+	vec2 pix_size = blur.pixel_size;
+	vec4 color = texture(source_color, uv_interp + vec2(-0.5, -0.5) * pix_size);
+	color += texture(source_color, uv_interp + vec2(0.5, -0.5) * pix_size);
+	color += texture(source_color, uv_interp + vec2(0.5, 0.5) * pix_size);
+	color += texture(source_color, uv_interp + vec2(-0.5, 0.5) * pix_size);
+	frag_color = color / 4.0;
+
+#endif
+
+#ifdef MODE_GAUSSIAN_BLUR
+
+	//Simpler blur uses SIGMA2 for the gaussian kernel for a stronger effect
+
+	if (bool(blur.flags & FLAG_HORIZONTAL)) {
+
+		vec2 pix_size = blur.pixel_size;
+		pix_size *= 0.5; //reading from larger buffer, so use more samples
+		vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.214607;
+		color += texture(source_color, uv_interp + vec2(1.0, 0.0) * pix_size) * 0.189879;
+		color += texture(source_color, uv_interp + vec2(2.0, 0.0) * pix_size) * 0.131514;
+		color += texture(source_color, uv_interp + vec2(3.0, 0.0) * pix_size) * 0.071303;
+		color += texture(source_color, uv_interp + vec2(-1.0, 0.0) * pix_size) * 0.189879;
+		color += texture(source_color, uv_interp + vec2(-2.0, 0.0) * pix_size) * 0.131514;
+		color += texture(source_color, uv_interp + vec2(-3.0, 0.0) * pix_size) * 0.071303;
+		frag_color = color;
+	} else {
+
+		vec2 pix_size = blur.pixel_size;
+		vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.38774;
+		color += texture(source_color, uv_interp + vec2(0.0, 1.0) * pix_size) * 0.24477;
+		color += texture(source_color, uv_interp + vec2(0.0, 2.0) * pix_size) * 0.06136;
+		color += texture(source_color, uv_interp + vec2(0.0, -1.0) * pix_size) * 0.24477;
+		color += texture(source_color, uv_interp + vec2(0.0, -2.0) * pix_size) * 0.06136;
+		frag_color = color;
+	}
+#endif
+
+#ifdef MODE_GAUSSIAN_GLOW
+
+	//Glow uses larger sigma 1 for a more rounded blur effect
+
+#define GLOW_ADD(m_ofs, m_mult)                                                  \
+	{                                                                            \
+		vec2 ofs = uv_interp + m_ofs * pix_size;                                 \
+		vec4 c = texture(source_color, ofs) * m_mult;                            \
+		if (any(lessThan(ofs, vec2(0.0))) || any(greaterThan(ofs, vec2(1.0)))) { \
+			c *= 0.0;                                                            \
+		}                                                                        \
+		color += c;                                                              \
+	}
+
+	if (bool(blur.flags & FLAG_HORIZONTAL)) {
+
+		vec2 pix_size = blur.pixel_size;
+		pix_size *= 0.5; //reading from larger buffer, so use more samples
+		vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.174938;
+		GLOW_ADD(vec2(1.0, 0.0), 0.165569);
+		GLOW_ADD(vec2(2.0, 0.0), 0.140367);
+		GLOW_ADD(vec2(3.0, 0.0), 0.106595);
+		GLOW_ADD(vec2(-1.0, 0.0), 0.165569);
+		GLOW_ADD(vec2(-2.0, 0.0), 0.140367);
+		GLOW_ADD(vec2(-3.0, 0.0), 0.106595);
+		color *= blur.glow_strength;
+		frag_color = color;
+	} else {
+
+		vec2 pix_size = blur.pixel_size;
+		vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.288713;
+		GLOW_ADD(vec2(0.0, 1.0), 0.233062);
+		GLOW_ADD(vec2(0.0, 2.0), 0.122581);
+		GLOW_ADD(vec2(0.0, -1.0), 0.233062);
+		GLOW_ADD(vec2(0.0, -2.0), 0.122581);
+		color *= blur.glow_strength;
+		frag_color = color;
+	}
+
+#undef GLOW_ADD
+
+	if (bool(blur.flags & FLAG_GLOW_FIRST_PASS)) {
+#ifdef GLOW_USE_AUTO_EXPOSURE
+
+		frag_color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / blur.glow_auto_exposure_grey;
+#endif
+		frag_color *= blur.glow_exposure;
+
+		float luminance = max(frag_color.r, max(frag_color.g, frag_color.b));
+		float feedback = max(smoothstep(blur.glow_hdr_threshold, blur.glow_hdr_threshold + blur.glow_hdr_scale, luminance), blur.glow_bloom);
+
+		frag_color = min(frag_color * feedback, vec4(blur.glow_luminance_cap));
+	}
+
+#endif
+
+#ifdef MODE_DOF_FAR_BLUR
+
+	vec4 color_accum = vec4(0.0);
+
+	float depth = texture(dof_source_depth, uv_interp, 0.0).r;
+	depth = depth * 2.0 - 1.0;
+
+	if (bool(blur.flags & FLAG_USE_ORTHOGONAL_PROJECTION)) {
+		depth = ((depth + (blur.camera_z_far + blur.camera_z_near) / (blur.camera_z_far - blur.camera_z_near)) * (blur.camera_z_far - blur.camera_z_near)) / 2.0;
+	} else {
+		depth = 2.0 * blur.camera_z_near * blur.camera_z_far / (blur.camera_z_far + blur.camera_z_near - depth * (blur.camera_z_far - blur.camera_z_near));
+	}
+
+	float amount = smoothstep(blur.dof_begin, blur.dof_end, depth);
+	float k_accum = 0.0;
+
+	for (int i = 0; i < dof_kernel_size; i++) {
+
+		int int_ofs = i - dof_kernel_from;
+		vec2 tap_uv = uv_interp + blur.dof_dir * float(int_ofs) * amount * blur.dof_radius;
+
+		float tap_k = dof_kernel[i];
+
+		float tap_depth = texture(dof_source_depth, tap_uv, 0.0).r;
+		tap_depth = tap_depth * 2.0 - 1.0;
+
+		if (bool(blur.flags & FLAG_USE_ORTHOGONAL_PROJECTION)) {
+
+			tap_depth = ((tap_depth + (blur.camera_z_far + blur.camera_z_near) / (blur.camera_z_far - blur.camera_z_near)) * (blur.camera_z_far - blur.camera_z_near)) / 2.0;
+		} else {
+			tap_depth = 2.0 * blur.camera_z_near * blur.camera_z_far / (blur.camera_z_far + blur.camera_z_near - tap_depth * (blur.camera_z_far - blur.camera_z_near));
+		}
+
+		float tap_amount = mix(smoothstep(blur.dof_begin, blur.dof_end, tap_depth), 1.0, int_ofs == 0);
+		tap_amount *= tap_amount * tap_amount; //prevent undesired glow effect
+
+		vec4 tap_color = texture(source_color, tap_uv, 0.0) * tap_k;
+
+		k_accum += tap_k * tap_amount;
+		color_accum += tap_color * tap_amount;
+	}
+
+	if (k_accum > 0.0) {
+		color_accum /= k_accum;
+	}
+
+	frag_color = color_accum; ///k_accum;
+
+#endif
+
+#ifdef MODE_DOF_NEAR_BLUR
+
+	vec4 color_accum = vec4(0.0);
+
+	float max_accum = 0.0;
+
+	for (int i = 0; i < dof_kernel_size; i++) {
+
+		int int_ofs = i - dof_kernel_from;
+		vec2 tap_uv = uv_interp + blur.dof_dir * float(int_ofs) * blur.dof_radius;
+		float ofs_influence = max(0.0, 1.0 - float(abs(int_ofs)) / float(dof_kernel_from));
+
+		float tap_k = dof_kernel[i];
+
+		vec4 tap_color = texture(source_color, tap_uv, 0.0);
+
+		float tap_depth = texture(dof_source_depth, tap_uv, 0.0).r;
+		tap_depth = tap_depth * 2.0 - 1.0;
+		if (bool(blur.flags & FLAG_USE_ORTHOGONAL_PROJECTION)) {
+
+			tap_depth = ((tap_depth + (blur.camera_z_far + blur.camera_z_near) / (blur.camera_z_far - blur.camera_z_near)) * (blur.camera_z_far - blur.camera_z_near)) / 2.0;
+		} else {
+			tap_depth = 2.0 * blur.camera_z_near * blur.camera_z_far / (blur.camera_z_far + blur.camera_z_near - tap_depth * (blur.camera_z_far - blur.camera_z_near));
+		}
+		float tap_amount = 1.0 - smoothstep(blur.dof_end, blur.dof_begin, tap_depth);
+		tap_amount *= tap_amount * tap_amount; //prevent undesired glow effect
+
+		if (bool(blur.flags & FLAG_DOF_NEAR_FIRST_TAP)) {
+			tap_color.a = 1.0 - smoothstep(blur.dof_end, blur.dof_begin, tap_depth);
+		}
+
+		max_accum = max(max_accum, tap_amount * ofs_influence);
+
+		color_accum += tap_color * tap_k;
+	}
+
+	color_accum.a = max(color_accum.a, sqrt(max_accum));
+
+#ifdef DOF_NEAR_BLUR_MERGE
+	{
+		vec4 original = texture(source_dof_original, uv_interp, 0.0);
+		color_accum = mix(original, color_accum, color_accum.a);
+	}
+#endif
+
+	if (bool(blur.flags & FLAG_DOF_NEAR_FIRST_TAP)) {
+		frag_color = color_accum;
+	}
+#endif
+
+#ifdef MODE_SIMPLE_COPY
+	vec4 color = texture(source_color, uv_interp, 0.0);
+	if (bool(blur.flags & FLAG_COPY_FORCE_LUMINANCE)) {
+		color.rgb = vec3(max(max(color.r, color.g), color.b));
+	}
+	frag_color = color;
+#endif
+
+#ifdef MODE_SSAO_MERGE
+	vec4 color = texture(source_color, uv_interp, 0.0);
+	float ssao = texture(source_ssao, uv_interp, 0.0).r;
+	frag_color = vec4(mix(color.rgb, color.rgb * mix(blur.ssao_color.rgb, vec3(1.0), ssao), color.a), 1.0);
+
+#endif
+}

servers/rendering/rasterizer_rd/shaders/blur_inc.glsl

@@ -0,0 +1,35 @@
+#define FLAG_HORIZONTAL (1 << 0)
+#define FLAG_USE_BLUR_SECTION (1 << 1)
+#define FLAG_USE_ORTHOGONAL_PROJECTION (1 << 2)
+#define FLAG_DOF_NEAR_FIRST_TAP (1 << 3)
+#define FLAG_GLOW_FIRST_PASS (1 << 4)
+#define FLAG_FLIP_Y (1 << 5)
+#define FLAG_COPY_FORCE_LUMINANCE (1 << 6)
+
+layout(push_constant, binding = 1, std430) uniform Blur {
+	vec4 section;
+	vec2 pixel_size;
+	uint flags;
+	uint pad;
+	// Glow.
+	float glow_strength;
+	float glow_bloom;
+	float glow_hdr_threshold;
+	float glow_hdr_scale;
+	float glow_exposure;
+	float glow_white;
+	float glow_luminance_cap;
+	float glow_auto_exposure_grey;
+	// DOF.
+	float dof_begin;
+	float dof_end;
+	float dof_radius;
+	float dof_pad;
+
+	vec2 dof_dir;
+	float camera_z_far;
+	float camera_z_near;
+
+	vec4 ssao_color;
+}
+blur;

servers/rendering/rasterizer_rd/shaders/bokeh_dof.glsl

@@ -0,0 +1,258 @@
+/* clang-format off */
+[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+#define BLOCK_SIZE 8
+
+layout(local_size_x = BLOCK_SIZE, local_size_y = BLOCK_SIZE, local_size_z = 1) in;
+/* clang-format on */
+
+#ifdef MODE_GEN_BLUR_SIZE
+layout(rgba16f, set = 0, binding = 0) uniform restrict image2D color_image;
+layout(set = 1, binding = 0) uniform sampler2D source_depth;
+#endif
+
+#if defined(MODE_BOKEH_BOX) || defined(MODE_BOKEH_HEXAGONAL) || defined(MODE_BOKEH_CIRCULAR)
+layout(set = 1, binding = 0) uniform sampler2D color_texture;
+layout(rgba16f, set = 0, binding = 0) uniform restrict writeonly image2D bokeh_image;
+#endif
+
+#ifdef MODE_COMPOSITE_BOKEH
+layout(rgba16f, set = 0, binding = 0) uniform restrict image2D color_image;
+layout(set = 1, binding = 0) uniform sampler2D source_bokeh;
+#endif
+
+// based on https://www.shadertoy.com/view/Xd3GDl
+
+layout(push_constant, binding = 1, std430) uniform Params {
+	ivec2 size;
+	float z_far;
+	float z_near;
+
+	bool orthogonal;
+	float blur_size;
+	float blur_scale;
+	int blur_steps;
+
+	bool blur_near_active;
+	float blur_near_begin;
+	float blur_near_end;
+	bool blur_far_active;
+
+	float blur_far_begin;
+	float blur_far_end;
+	bool second_pass;
+	bool half_size;
+
+	bool use_jitter;
+	float jitter_seed;
+	uint pad[2];
+}
+params;
+
+//used to work around downsampling filter
+#define DEPTH_GAP 0.0
+
+#ifdef MODE_GEN_BLUR_SIZE
+
+float get_depth_at_pos(vec2 uv) {
+	float depth = textureLod(source_depth, uv, 0.0).x;
+	if (params.orthogonal) {
+		depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0;
+	} else {
+		depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near));
+	}
+	return depth;
+}
+
+float get_blur_size(float depth) {
+
+	if (params.blur_near_active && depth < params.blur_near_begin) {
+		return -(1.0 - smoothstep(params.blur_near_end, params.blur_near_begin, depth)) * params.blur_size - DEPTH_GAP; //near blur is negative
+	}
+
+	if (params.blur_far_active && depth > params.blur_far_begin) {
+		return smoothstep(params.blur_far_begin, params.blur_far_end, depth) * params.blur_size + DEPTH_GAP;
+	}
+
+	return 0.0;
+}
+
+#endif
+
+const float GOLDEN_ANGLE = 2.39996323;
+
+//note: uniform pdf rand [0;1[
+float hash12n(vec2 p) {
+	p = fract(p * vec2(5.3987, 5.4421));
+	p += dot(p.yx, p.xy + vec2(21.5351, 14.3137));
+	return fract(p.x * p.y * 95.4307);
+}
+
+#if defined(MODE_BOKEH_BOX) || defined(MODE_BOKEH_HEXAGONAL)
+
+vec4 weighted_filter_dir(vec2 dir, vec2 uv, vec2 pixel_size) {
+
+	dir *= pixel_size;
+	vec4 color = texture(color_texture, uv);
+
+	vec4 accum = color;
+	float total = 1.0;
+
+	float blur_scale = params.blur_size / float(params.blur_steps);
+
+	if (params.use_jitter) {
+		uv += dir * (hash12n(uv + params.jitter_seed) - 0.5);
+	}
+
+	for (int i = -params.blur_steps; i <= params.blur_steps; i++) {
+
+		if (i == 0) {
+			continue;
+		}
+		float radius = float(i) * blur_scale;
+		vec2 suv = uv + dir * radius;
+		radius = abs(radius);
+
+		vec4 sample_color = texture(color_texture, suv);
+		float limit;
+
+		if (sample_color.a < color.a) {
+			limit = abs(sample_color.a);
+		} else {
+			limit = abs(color.a);
+		}
+
+		limit -= DEPTH_GAP;
+
+		float m = smoothstep(radius - 0.5, radius + 0.5, limit);
+
+		accum += mix(color, sample_color, m);
+
+		total += 1.0;
+	}
+
+	return accum / total;
+}
+
+#endif
+
+void main() {
+
+	ivec2 pos = ivec2(gl_GlobalInvocationID.xy);
+
+	if (any(greaterThan(pos, params.size))) { //too large, do nothing
+		return;
+	}
+
+	vec2 pixel_size = 1.0 / vec2(params.size);
+	vec2 uv = vec2(pos) / vec2(params.size);
+
+#ifdef MODE_GEN_BLUR_SIZE
+	uv += pixel_size * 0.5;
+	//precompute size in alpha channel
+	float depth = get_depth_at_pos(uv);
+	float size = get_blur_size(depth);
+
+	vec4 color = imageLoad(color_image, pos);
+	color.a = size;
+	imageStore(color_image, pos, color);
+#endif
+
+#ifdef MODE_BOKEH_BOX
+
+	//pixel_size*=0.5; //resolution is doubled
+	if (params.second_pass || !params.half_size) {
+		uv += pixel_size * 0.5; //half pixel to read centers
+	} else {
+		uv += pixel_size * 0.25; //half pixel to read centers from full res
+	}
+
+	vec2 dir = (params.second_pass ? vec2(0.0, 1.0) : vec2(1.0, 0.0));
+
+	vec4 color = weighted_filter_dir(dir, uv, pixel_size);
+
+	imageStore(bokeh_image, pos, color);
+
+#endif
+
+#ifdef MODE_BOKEH_HEXAGONAL
+
+	//pixel_size*=0.5; //resolution is doubled
+	if (params.second_pass || !params.half_size) {
+		uv += pixel_size * 0.5; //half pixel to read centers
+	} else {
+		uv += pixel_size * 0.25; //half pixel to read centers from full res
+	}
+
+	vec2 dir = (params.second_pass ? normalize(vec2(1.0, 0.577350269189626)) : vec2(0.0, 1.0));
+
+	vec4 color = weighted_filter_dir(dir, uv, pixel_size);
+
+	if (params.second_pass) {
+		dir = normalize(vec2(-1.0, 0.577350269189626));
+
+		vec4 color2 = weighted_filter_dir(dir, uv, pixel_size);
+
+		color.rgb = min(color.rgb, color2.rgb);
+		color.a = (color.a + color2.a) * 0.5;
+	}
+
+	imageStore(bokeh_image, pos, color);
+
+#endif
+
+#ifdef MODE_BOKEH_CIRCULAR
+
+	if (params.half_size) {
+		pixel_size *= 0.5; //resolution is doubled
+	}
+
+	uv += pixel_size * 0.5; //half pixel to read centers
+
+	vec4 color = texture(color_texture, uv);
+	float accum = 1.0;
+	float radius = params.blur_scale;
+
+	for (float ang = 0.0; radius < params.blur_size; ang += GOLDEN_ANGLE) {
+
+		vec2 suv = uv + vec2(cos(ang), sin(ang)) * pixel_size * radius;
+		vec4 sample_color = texture(color_texture, suv);
+		float sample_size = abs(sample_color.a);
+		if (sample_color.a > color.a) {
+			sample_size = clamp(sample_size, 0.0, abs(color.a) * 2.0);
+		}
+
+		float m = smoothstep(radius - 0.5, radius + 0.5, sample_size);
+		color += mix(color / accum, sample_color, m);
+		accum += 1.0;
+		radius += params.blur_scale / radius;
+	}
+
+	color /= accum;
+
+	imageStore(bokeh_image, pos, color);
+#endif
+
+#ifdef MODE_COMPOSITE_BOKEH
+
+	uv += pixel_size * 0.5;
+	vec4 color = imageLoad(color_image, pos);
+	vec4 bokeh = texture(source_bokeh, uv);
+
+	float mix_amount;
+	if (bokeh.a < color.a) {
+		mix_amount = min(1.0, max(0.0, max(abs(color.a), abs(bokeh.a)) - DEPTH_GAP));
+	} else {
+		mix_amount = min(1.0, max(0.0, abs(color.a) - DEPTH_GAP));
+	}
+
+	color.rgb = mix(color.rgb, bokeh.rgb, mix_amount); //blend between hires and lowres
+
+	color.a = 0; //reset alpha
+	imageStore(color_image, pos, color);
+#endif
+}

servers/rendering/rasterizer_rd/shaders/canvas.glsl

@@ -0,0 +1,584 @@
+/* clang-format off */
+[vertex]
+
+#version 450
+
+VERSION_DEFINES
+
+#ifdef USE_ATTRIBUTES
+layout(location = 0) in vec2 vertex_attrib;
+/* clang-format on */
+layout(location = 3) in vec4 color_attrib;
+layout(location = 4) in vec2 uv_attrib;
+
+layout(location = 6) in uvec4 bones_attrib;
+
+#endif
+
+#include "canvas_uniforms_inc.glsl"
+
+layout(location = 0) out vec2 uv_interp;
+layout(location = 1) out vec4 color_interp;
+layout(location = 2) out vec2 vertex_interp;
+
+#ifdef USE_NINEPATCH
+
+layout(location = 3) out vec2 pixel_size_interp;
+
+#endif
+
+#ifdef USE_MATERIAL_UNIFORMS
+layout(set = 1, binding = 1, std140) uniform MaterialUniforms{
+	/* clang-format off */
+MATERIAL_UNIFORMS
+	/* clang-format on */
+} material;
+#endif
+
+/* clang-format off */
+VERTEX_SHADER_GLOBALS
+/* clang-format on */
+
+void main() {
+
+	vec4 instance_custom = vec4(0.0);
+#ifdef USE_PRIMITIVE
+
+	//weird bug,
+	//this works
+	vec2 vertex;
+	vec2 uv;
+	vec4 color;
+
+	if (gl_VertexIndex == 0) {
+		vertex = draw_data.points[0];
+		uv = draw_data.uvs[0];
+		color = vec4(unpackHalf2x16(draw_data.colors[0]), unpackHalf2x16(draw_data.colors[1]));
+	} else if (gl_VertexIndex == 1) {
+		vertex = draw_data.points[1];
+		uv = draw_data.uvs[1];
+		color = vec4(unpackHalf2x16(draw_data.colors[2]), unpackHalf2x16(draw_data.colors[3]));
+	} else {
+		vertex = draw_data.points[2];
+		uv = draw_data.uvs[2];
+		color = vec4(unpackHalf2x16(draw_data.colors[4]), unpackHalf2x16(draw_data.colors[5]));
+	}
+	uvec4 bones = uvec4(0, 0, 0, 0);
+
+#elif defined(USE_ATTRIBUTES)
+
+	vec2 vertex = vertex_attrib;
+	vec4 color = color_attrib;
+	vec2 uv = uv_attrib;
+
+	uvec4 bones = bones_attrib;
+#else
+
+	vec2 vertex_base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0));
+	vec2 vertex_base = vertex_base_arr[gl_VertexIndex];
+
+	vec2 uv = draw_data.src_rect.xy + abs(draw_data.src_rect.zw) * ((draw_data.flags & FLAGS_TRANSPOSE_RECT) != 0 ? vertex_base.yx : vertex_base.xy);
+	vec4 color = draw_data.modulation;
+	vec2 vertex = draw_data.dst_rect.xy + abs(draw_data.dst_rect.zw) * mix(vertex_base, vec2(1.0, 1.0) - vertex_base, lessThan(draw_data.src_rect.zw, vec2(0.0, 0.0)));
+	uvec4 bones = uvec4(0, 0, 0, 0);
+
+#endif
+
+	mat4 world_matrix = mat4(vec4(draw_data.world_x, 0.0, 0.0), vec4(draw_data.world_y, 0.0, 0.0), vec4(0.0, 0.0, 1.0, 0.0), vec4(draw_data.world_ofs, 0.0, 1.0));
+
+#if 0
+	if (draw_data.flags & FLAGS_INSTANCING_ENABLED) {
+
+		uint offset = draw_data.flags & FLAGS_INSTANCING_STRIDE_MASK;
+		offset *= gl_InstanceIndex;
+		mat4 instance_xform = mat4(
+				vec4(texelFetch(instancing_buffer, offset + 0), texelFetch(instancing_buffer, offset + 1), 0.0, texelFetch(instancing_buffer, offset + 3)),
+				vec4(texelFetch(instancing_buffer, offset + 4), texelFetch(instancing_buffer, offset + 5), 0.0, texelFetch(instancing_buffer, offset + 7)),
+				vec4(0.0, 0.0, 1.0, 0.0),
+				vec4(0.0, 0.0, 0.0, 1.0));
+		offset += 8;
+		if (draw_data.flags & FLAGS_INSTANCING_HAS_COLORS) {
+			vec4 instance_color;
+			if (draw_data.flags & FLAGS_INSTANCING_COLOR_8_BIT) {
+				uint bits = floatBitsToUint(texelFetch(instancing_buffer, offset));
+				instance_color = unpackUnorm4x8(bits);
+				offset += 1;
+			} else {
+				instance_color = vec4(texelFetch(instancing_buffer, offset + 0), texelFetch(instancing_buffer, offset + 1), texelFetch(instancing_buffer, offset + 2), texelFetch(instancing_buffer, offset + 3));
+				offser += 4;
+			}
+
+			color *= instance_color;
+		}
+		if (draw_data.flags & FLAGS_INSTANCING_HAS_CUSTOM_DATA) {
+			if (draw_data.flags & FLAGS_INSTANCING_CUSTOM_DATA_8_BIT) {
+				uint bits = floatBitsToUint(texelFetch(instancing_buffer, offset));
+				instance_custom = unpackUnorm4x8(bits);
+			} else {
+				instance_custom = vec4(texelFetch(instancing_buffer, offset + 0), texelFetch(instancing_buffer, offset + 1), texelFetch(instancing_buffer, offset + 2), texelFetch(instancing_buffer, offset + 3));
+			}
+		}
+	}
+
+#endif
+
+#if !defined(USE_ATTRIBUTES) && !defined(USE_PRIMITIVE)
+	if (bool(draw_data.flags & FLAGS_USING_PARTICLES)) {
+		//scale by texture size
+		vertex /= draw_data.color_texture_pixel_size;
+	}
+#endif
+
+#ifdef USE_POINT_SIZE
+	float point_size = 1.0;
+#endif
+	{
+		/* clang-format off */
+VERTEX_SHADER_CODE
+		/* clang-format on */
+	}
+
+#ifdef USE_NINEPATCH
+	pixel_size_interp = abs(draw_data.dst_rect.zw) * vertex_base;
+#endif
+
+#if !defined(SKIP_TRANSFORM_USED)
+	vertex = (world_matrix * vec4(vertex, 0.0, 1.0)).xy;
+#endif
+
+	color_interp = color;
+
+	if (bool(draw_data.flags & FLAGS_USE_PIXEL_SNAP)) {
+
+		vertex = floor(vertex + 0.5);
+		// precision issue on some hardware creates artifacts within texture
+		// offset uv by a small amount to avoid
+		uv += 1e-5;
+	}
+
+#ifdef USE_ATTRIBUTES
+#if 0
+	if (bool(draw_data.flags & FLAGS_USE_SKELETON) && bone_weights != vec4(0.0)) { //must be a valid bone
+		//skeleton transform
+
+		ivec4 bone_indicesi = ivec4(bone_indices);
+
+		uvec2 tex_ofs = bone_indicesi.x * 2;
+
+		mat2x4 m;
+		m = mat2x4(
+					texelFetch(skeleton_buffer, tex_ofs + 0),
+					texelFetch(skeleton_buffer, tex_ofs + 1)) *
+			bone_weights.x;
+
+		tex_ofs = bone_indicesi.y * 2;
+
+		m += mat2x4(
+					 texelFetch(skeleton_buffer, tex_ofs + 0),
+					 texelFetch(skeleton_buffer, tex_ofs + 1)) *
+			 bone_weights.y;
+
+		tex_ofs = bone_indicesi.z * 2;
+
+		m += mat2x4(
+					 texelFetch(skeleton_buffer, tex_ofs + 0),
+					 texelFetch(skeleton_buffer, tex_ofs + 1)) *
+			 bone_weights.z;
+
+		tex_ofs = bone_indicesi.w * 2;
+
+		m += mat2x4(
+					 texelFetch(skeleton_buffer, tex_ofs + 0),
+					 texelFetch(skeleton_buffer, tex_ofs + 1)) *
+			 bone_weights.w;
+
+		mat4 bone_matrix = skeleton_data.skeleton_transform * transpose(mat4(m[0], m[1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))) * skeleton_data.skeleton_transform_inverse;
+
+		//outvec = bone_matrix * outvec;
+	}
+#endif
+#endif
+
+	vertex = (canvas_data.canvas_transform * vec4(vertex, 0.0, 1.0)).xy;
+
+	vertex_interp = vertex;
+	uv_interp = uv;
+
+	gl_Position = canvas_data.screen_transform * vec4(vertex, 0.0, 1.0);
+
+#ifdef USE_POINT_SIZE
+	gl_PointSize = point_size;
+#endif
+}
+
+/* clang-format off */
+[fragment]
+
+#version 450
+
+VERSION_DEFINES
+
+#include "canvas_uniforms_inc.glsl"
+
+layout(location = 0) in vec2 uv_interp;
+/* clang-format on */
+layout(location = 1) in vec4 color_interp;
+layout(location = 2) in vec2 vertex_interp;
+
+#ifdef USE_NINEPATCH
+
+layout(location = 3) in vec2 pixel_size_interp;
+
+#endif
+
+layout(location = 0) out vec4 frag_color;
+
+#ifdef USE_MATERIAL_UNIFORMS
+layout(set = 1, binding = 1, std140) uniform MaterialUniforms{
+	/* clang-format off */
+MATERIAL_UNIFORMS
+	/* clang-format on */
+} material;
+#endif
+
+/* clang-format off */
+FRAGMENT_SHADER_GLOBALS
+/* clang-format on */
+
+#ifdef LIGHT_SHADER_CODE_USED
+
+vec4 light_compute(
+		vec3 light_vertex,
+		vec3 light_position,
+		vec3 normal,
+		vec4 light_color,
+		float light_energy,
+		vec4 specular_shininess,
+		inout vec4 shadow_modulate,
+		vec2 screen_uv,
+		vec2 uv,
+		vec4 color) {
+
+	vec4 light = vec4(0.0);
+	/* clang-format off */
+LIGHT_SHADER_CODE
+	/* clang-format on */
+	return light;
+}
+
+#endif
+
+#ifdef USE_NINEPATCH
+
+float map_ninepatch_axis(float pixel, float draw_size, float tex_pixel_size, float margin_begin, float margin_end, int np_repeat, inout int draw_center) {
+
+	float tex_size = 1.0 / tex_pixel_size;
+
+	if (pixel < margin_begin) {
+		return pixel * tex_pixel_size;
+	} else if (pixel >= draw_size - margin_end) {
+		return (tex_size - (draw_size - pixel)) * tex_pixel_size;
+	} else {
+		if (!bool(draw_data.flags & FLAGS_NINEPACH_DRAW_CENTER)) {
+			draw_center--;
+		}
+
+		// np_repeat is passed as uniform using NinePatchRect::AxisStretchMode enum.
+		if (np_repeat == 0) { // Stretch.
+			// Convert to ratio.
+			float ratio = (pixel - margin_begin) / (draw_size - margin_begin - margin_end);
+			// Scale to source texture.
+			return (margin_begin + ratio * (tex_size - margin_begin - margin_end)) * tex_pixel_size;
+		} else if (np_repeat == 1) { // Tile.
+			// Convert to offset.
+			float ofs = mod((pixel - margin_begin), tex_size - margin_begin - margin_end);
+			// Scale to source texture.
+			return (margin_begin + ofs) * tex_pixel_size;
+		} else if (np_repeat == 2) { // Tile Fit.
+			// Calculate scale.
+			float src_area = draw_size - margin_begin - margin_end;
+			float dst_area = tex_size - margin_begin - margin_end;
+			float scale = max(1.0, floor(src_area / max(dst_area, 0.0000001) + 0.5));
+			// Convert to ratio.
+			float ratio = (pixel - margin_begin) / src_area;
+			ratio = mod(ratio * scale, 1.0);
+			// Scale to source texture.
+			return (margin_begin + ratio * dst_area) * tex_pixel_size;
+		} else { // Shouldn't happen, but silences compiler warning.
+			return 0.0;
+		}
+	}
+}
+
+#endif
+
+void main() {
+
+	vec4 color = color_interp;
+	vec2 uv = uv_interp;
+	vec2 vertex = vertex_interp;
+
+#if !defined(USE_ATTRIBUTES) && !defined(USE_PRIMITIVE)
+
+#ifdef USE_NINEPATCH
+
+	int draw_center = 2;
+	uv = vec2(
+			map_ninepatch_axis(pixel_size_interp.x, abs(draw_data.dst_rect.z), draw_data.color_texture_pixel_size.x, draw_data.ninepatch_margins.x, draw_data.ninepatch_margins.z, int(draw_data.flags >> FLAGS_NINEPATCH_H_MODE_SHIFT) & 0x3, draw_center),
+			map_ninepatch_axis(pixel_size_interp.y, abs(draw_data.dst_rect.w), draw_data.color_texture_pixel_size.y, draw_data.ninepatch_margins.y, draw_data.ninepatch_margins.w, int(draw_data.flags >> FLAGS_NINEPATCH_V_MODE_SHIFT) & 0x3, draw_center));
+
+	if (draw_center == 0) {
+		color.a = 0.0;
+	}
+
+	uv = uv * draw_data.src_rect.zw + draw_data.src_rect.xy; //apply region if needed
+
+#endif
+	if (bool(draw_data.flags & FLAGS_CLIP_RECT_UV)) {
+
+		uv = clamp(uv, draw_data.src_rect.xy, draw_data.src_rect.xy + abs(draw_data.src_rect.zw));
+	}
+
+#endif
+
+	color *= texture(sampler2D(color_texture, texture_sampler), uv);
+
+	uint light_count = (draw_data.flags >> FLAGS_LIGHT_COUNT_SHIFT) & 0xF; //max 16 lights
+
+	vec3 normal;
+
+#if defined(NORMAL_USED)
+
+	bool normal_used = true;
+#else
+	bool normal_used = false;
+#endif
+
+	if (normal_used || (light_count > 0 && bool(draw_data.flags & FLAGS_DEFAULT_NORMAL_MAP_USED))) {
+		normal.xy = texture(sampler2D(normal_texture, texture_sampler), uv).xy * vec2(2.0, -2.0) - vec2(1.0, -1.0);
+		normal.z = sqrt(1.0 - dot(normal.xy, normal.xy));
+		normal_used = true;
+	} else {
+		normal = vec3(0.0, 0.0, 1.0);
+	}
+
+	vec4 specular_shininess;
+
+#if defined(SPECULAR_SHININESS_USED)
+
+	bool specular_shininess_used = true;
+#else
+	bool specular_shininess_used = false;
+#endif
+
+	if (specular_shininess_used || (light_count > 0 && normal_used && bool(draw_data.flags & FLAGS_DEFAULT_SPECULAR_MAP_USED))) {
+		specular_shininess = texture(sampler2D(specular_texture, texture_sampler), uv);
+		specular_shininess *= unpackUnorm4x8(draw_data.specular_shininess);
+		specular_shininess_used = true;
+	} else {
+		specular_shininess = vec4(1.0);
+	}
+
+#if defined(SCREEN_UV_USED)
+	vec2 screen_uv = gl_FragCoord.xy * canvas_data.screen_pixel_size;
+#else
+	vec2 screen_uv = vec2(0.0);
+#endif
+
+	vec3 light_vertex = vec3(vertex, 0.0);
+	vec2 shadow_vertex = vertex;
+
+	{
+		float normal_depth = 1.0;
+
+#if defined(NORMALMAP_USED)
+		vec3 normal_map = vec3(0.0, 0.0, 1.0);
+		normal_used = true;
+#endif
+
+		/* clang-format off */
+
+FRAGMENT_SHADER_CODE
+
+		/* clang-format on */
+
+#if defined(NORMALMAP_USED)
+		normal = mix(vec3(0.0, 0.0, 1.0), normal_map * vec3(2.0, -2.0, 1.0) - vec3(1.0, -1.0, 0.0), normal_depth);
+#endif
+	}
+
+	if (normal_used) {
+		//convert by item transform
+		normal.xy = mat2(normalize(draw_data.world_x), normalize(draw_data.world_y)) * normal.xy;
+		//convert by canvas transform
+		normal = normalize((canvas_data.canvas_normal_transform * vec4(normal, 0.0)).xyz);
+	}
+
+	vec4 base_color = color;
+	if (bool(draw_data.flags & FLAGS_USING_LIGHT_MASK)) {
+		color = vec4(0.0); //invisible by default due to using light mask
+	}
+
+	color *= canvas_data.canvas_modulation;
+#ifdef USE_LIGHTING
+	for (uint i = 0; i < MAX_LIGHT_TEXTURES; i++) {
+		if (i >= light_count) {
+			break;
+		}
+		uint light_base;
+		if (i < 8) {
+			if (i < 4) {
+				light_base = draw_data.lights[0];
+			} else {
+				light_base = draw_data.lights[1];
+			}
+		} else {
+			if (i < 12) {
+				light_base = draw_data.lights[2];
+			} else {
+				light_base = draw_data.lights[3];
+			}
+		}
+		light_base >>= (i & 3) * 8;
+		light_base &= 0xFF;
+
+		vec2 tex_uv = (vec4(vertex, 0.0, 1.0) * mat4(light_array.data[light_base].texture_matrix[0], light_array.data[light_base].texture_matrix[1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy; //multiply inverse given its transposed. Optimizer removes useless operations.
+		vec4 light_color = texture(sampler2D(light_textures[i], texture_sampler), tex_uv);
+		vec4 light_base_color = light_array.data[light_base].color;
+
+#ifdef LIGHT_SHADER_CODE_USED
+
+		vec4 shadow_modulate = vec4(1.0);
+		vec3 light_position = vec3(light_array.data[light_base].position, light_array.data[light_base].height);
+
+		light_color.rgb *= light_base_color.rgb;
+		light_color = light_compute(light_vertex, light_position, normal, light_color, light_base_color.a, specular_shininess, shadow_modulate, screen_uv, color, uv);
+#else
+
+		light_color.rgb *= light_base_color.rgb * light_base_color.a;
+
+		if (normal_used) {
+
+			vec3 light_pos = vec3(light_array.data[light_base].position, light_array.data[light_base].height);
+			vec3 pos = light_vertex;
+			vec3 light_vec = normalize(light_pos - pos);
+			float cNdotL = max(0.0, dot(normal, light_vec));
+
+			if (specular_shininess_used) {
+				//blinn
+				vec3 view = vec3(0.0, 0.0, 1.0); // not great but good enough
+				vec3 half_vec = normalize(view + light_vec);
+
+				float cNdotV = max(dot(normal, view), 0.0);
+				float cNdotH = max(dot(normal, half_vec), 0.0);
+				float cVdotH = max(dot(view, half_vec), 0.0);
+				float cLdotH = max(dot(light_vec, half_vec), 0.0);
+				float shininess = exp2(15.0 * specular_shininess.a + 1.0) * 0.25;
+				float blinn = pow(cNdotH, shininess);
+				blinn *= (shininess + 8.0) * (1.0 / (8.0 * M_PI));
+				float s = (blinn) / max(4.0 * cNdotV * cNdotL, 0.75);
+
+				light_color.rgb = specular_shininess.rgb * light_base_color.rgb * s + light_color.rgb * cNdotL;
+			} else {
+				light_color.rgb *= cNdotL;
+			}
+		}
+#endif
+		if (any(lessThan(tex_uv, vec2(0.0, 0.0))) || any(greaterThanEqual(tex_uv, vec2(1.0, 1.0)))) {
+			//if outside the light texture, light color is zero
+			light_color.a = 0.0;
+		}
+
+		if (bool(light_array.data[light_base].flags & LIGHT_FLAGS_HAS_SHADOW)) {
+
+			vec2 shadow_pos = (vec4(shadow_vertex, 0.0, 1.0) * mat4(light_array.data[light_base].shadow_matrix[0], light_array.data[light_base].shadow_matrix[1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy; //multiply inverse given its transposed. Optimizer removes useless operations.
+
+			vec2 pos_norm = normalize(shadow_pos);
+			vec2 pos_abs = abs(pos_norm);
+			vec2 pos_box = pos_norm / max(pos_abs.x, pos_abs.y);
+			vec2 pos_rot = pos_norm * mat2(vec2(0.7071067811865476, -0.7071067811865476), vec2(0.7071067811865476, 0.7071067811865476)); //is there a faster way to 45 degrees rot?
+			float tex_ofs;
+			float distance;
+			if (pos_rot.y > 0) {
+				if (pos_rot.x > 0) {
+					tex_ofs = pos_box.y * 0.125 + 0.125;
+					distance = shadow_pos.x;
+				} else {
+					tex_ofs = pos_box.x * -0.125 + (0.25 + 0.125);
+					distance = shadow_pos.y;
+				}
+			} else {
+				if (pos_rot.x < 0) {
+					tex_ofs = pos_box.y * -0.125 + (0.5 + 0.125);
+					distance = -shadow_pos.x;
+				} else {
+					tex_ofs = pos_box.x * 0.125 + (0.75 + 0.125);
+					distance = -shadow_pos.y;
+				}
+			}
+
+			//float distance = length(shadow_pos);
+			float shadow;
+			uint shadow_mode = light_array.data[light_base].flags & LIGHT_FLAGS_FILTER_MASK;
+
+			vec4 shadow_uv = vec4(tex_ofs, 0.0, distance, 1.0);
+
+			if (shadow_mode == LIGHT_FLAGS_SHADOW_NEAREST) {
+				shadow = textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv).x;
+			} else if (shadow_mode == LIGHT_FLAGS_SHADOW_PCF5) {
+				vec4 shadow_pixel_size = vec4(light_array.data[light_base].shadow_pixel_size, 0.0, 0.0, 0.0);
+				shadow = 0.0;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv - shadow_pixel_size * 2.0).x;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv - shadow_pixel_size).x;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv).x;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv + shadow_pixel_size).x;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv + shadow_pixel_size * 2.0).x;
+				shadow /= 5.0;
+			} else { //PCF13
+				vec4 shadow_pixel_size = vec4(light_array.data[light_base].shadow_pixel_size, 0.0, 0.0, 0.0);
+				shadow = 0.0;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv - shadow_pixel_size * 6.0).x;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv - shadow_pixel_size * 5.0).x;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv - shadow_pixel_size * 4.0).x;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv - shadow_pixel_size * 3.0).x;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv - shadow_pixel_size * 2.0).x;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv - shadow_pixel_size).x;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv).x;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv + shadow_pixel_size).x;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv + shadow_pixel_size * 2.0).x;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv + shadow_pixel_size * 3.0).x;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv + shadow_pixel_size * 4.0).x;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv + shadow_pixel_size * 5.0).x;
+				shadow += textureProj(sampler2DShadow(shadow_textures[i], shadow_sampler), shadow_uv + shadow_pixel_size * 6.0).x;
+				shadow /= 13.0;
+			}
+
+			vec4 shadow_color = light_array.data[light_base].shadow_color;
+#ifdef LIGHT_SHADER_CODE_USED
+			shadow_color *= shadow_modulate;
+#endif
+			light_color = mix(light_color, shadow_color, shadow);
+		}
+
+		uint blend_mode = light_array.data[light_base].flags & LIGHT_FLAGS_BLEND_MASK;
+
+		switch (blend_mode) {
+			case LIGHT_FLAGS_BLEND_MODE_ADD: {
+				color.rgb += light_color.rgb * light_color.a;
+			} break;
+			case LIGHT_FLAGS_BLEND_MODE_SUB: {
+				color.rgb -= light_color.rgb * light_color.a;
+			} break;
+			case LIGHT_FLAGS_BLEND_MODE_MIX: {
+				color.rgb = mix(color.rgb, light_color.rgb, light_color.a);
+			} break;
+			case LIGHT_FLAGS_BLEND_MODE_MASK: {
+				light_color.a *= base_color.a;
+				color.rgb = mix(color.rgb, light_color.rgb, light_color.a);
+			} break;
+		}
+	}
+#endif
+
+	frag_color = color;
+}

servers/rendering/rasterizer_rd/shaders/canvas_occlusion.glsl

@@ -0,0 +1,40 @@
+/* clang-format off */
+[vertex]
+
+#version 450
+
+layout(location = 0) in highp vec3 vertex;
+/* clang-format on */
+
+layout(push_constant, binding = 0, std430) uniform Constants {
+
+	mat4 projection;
+	mat2x4 modelview;
+	vec2 direction;
+	vec2 pad;
+}
+constants;
+
+layout(location = 0) out highp float depth;
+
+void main() {
+
+	highp vec4 vtx = vec4(vertex, 1.0) * mat4(constants.modelview[0], constants.modelview[1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0));
+	depth = dot(constants.direction, vtx.xy);
+
+	gl_Position = constants.projection * vtx;
+}
+
+/* clang-format off */
+[fragment]
+
+#version 450
+
+layout(location = 0) in highp float depth;
+/* clang-format on */
+layout(location = 0) out highp float distance_buf;
+
+void main() {
+
+	distance_buf = depth;
+}

servers/rendering/rasterizer_rd/shaders/canvas_uniforms_inc.glsl

@@ -0,0 +1,141 @@
+#define M_PI 3.14159265359
+
+#define FLAGS_INSTANCING_STRIDE_MASK 0xF
+#define FLAGS_INSTANCING_ENABLED (1 << 4)
+#define FLAGS_INSTANCING_HAS_COLORS (1 << 5)
+#define FLAGS_INSTANCING_COLOR_8BIT (1 << 6)
+#define FLAGS_INSTANCING_HAS_CUSTOM_DATA (1 << 7)
+#define FLAGS_INSTANCING_CUSTOM_DATA_8_BIT (1 << 8)
+
+#define FLAGS_CLIP_RECT_UV (1 << 9)
+#define FLAGS_TRANSPOSE_RECT (1 << 10)
+#define FLAGS_USING_LIGHT_MASK (1 << 11)
+#define FLAGS_NINEPACH_DRAW_CENTER (1 << 12)
+#define FLAGS_USING_PARTICLES (1 << 13)
+#define FLAGS_USE_PIXEL_SNAP (1 << 14)
+
+#define FLAGS_NINEPATCH_H_MODE_SHIFT 16
+#define FLAGS_NINEPATCH_V_MODE_SHIFT 18
+
+#define FLAGS_LIGHT_COUNT_SHIFT 20
+
+#define FLAGS_DEFAULT_NORMAL_MAP_USED (1 << 26)
+#define FLAGS_DEFAULT_SPECULAR_MAP_USED (1 << 27)
+
+// In vulkan, sets should always be ordered using the following logic:
+// Lower Sets: Sets that change format and layout less often
+// Higher sets: Sets that change format and layout very often
+// This is because changing a set for another with a different layout or format,
+// invalidates all the upper ones.
+
+/* SET0: Draw Primitive */
+
+layout(push_constant, binding = 0, std430) uniform DrawData {
+	vec2 world_x;
+	vec2 world_y;
+	vec2 world_ofs;
+	uint flags;
+	uint specular_shininess;
+#ifdef USE_PRIMITIVE
+	vec2 points[3];
+	vec2 uvs[3];
+	uint colors[6];
+#else
+	vec4 modulation;
+	vec4 ninepatch_margins;
+	vec4 dst_rect; //for built-in rect and UV
+	vec4 src_rect;
+	vec2 pad;
+
+#endif
+	vec2 color_texture_pixel_size;
+	uint lights[4];
+}
+draw_data;
+
+// The values passed per draw primitives are cached within it
+
+layout(set = 0, binding = 1) uniform texture2D color_texture;
+layout(set = 0, binding = 2) uniform texture2D normal_texture;
+layout(set = 0, binding = 3) uniform texture2D specular_texture;
+layout(set = 0, binding = 4) uniform sampler texture_sampler;
+
+layout(set = 0, binding = 5) uniform textureBuffer instancing_buffer;
+
+/* SET1: Is reserved for the material */
+
+#ifdef USE_MATERIAL_SAMPLERS
+
+layout(set = 1, binding = 0) uniform sampler material_samplers[12];
+
+#endif
+
+/* SET2: Canvas Item State (including lighting) */
+
+layout(set = 2, binding = 0, std140) uniform CanvasData {
+	mat4 canvas_transform;
+	mat4 screen_transform;
+	mat4 canvas_normal_transform;
+	vec4 canvas_modulation;
+	vec2 screen_pixel_size;
+	float time;
+	float time_pad;
+	//uint light_count;
+}
+canvas_data;
+
+layout(set = 2, binding = 1) uniform textureBuffer skeleton_buffer;
+
+layout(set = 2, binding = 2, std140) uniform SkeletonData {
+	mat4 skeleton_transform; //in world coordinates
+	mat4 skeleton_transform_inverse;
+}
+skeleton_data;
+
+#ifdef USE_LIGHTING
+
+#define LIGHT_FLAGS_BLEND_MASK (3 << 16)
+#define LIGHT_FLAGS_BLEND_MODE_ADD (0 << 16)
+#define LIGHT_FLAGS_BLEND_MODE_SUB (1 << 16)
+#define LIGHT_FLAGS_BLEND_MODE_MIX (2 << 16)
+#define LIGHT_FLAGS_BLEND_MODE_MASK (3 << 16)
+#define LIGHT_FLAGS_HAS_SHADOW (1 << 20)
+#define LIGHT_FLAGS_FILTER_SHIFT 22
+#define LIGHT_FLAGS_FILTER_MASK (3 << 22)
+#define LIGHT_FLAGS_SHADOW_NEAREST (0 << 22)
+#define LIGHT_FLAGS_SHADOW_PCF5 (1 << 22)
+#define LIGHT_FLAGS_SHADOW_PCF13 (2 << 22)
+
+struct Light {
+	mat2x4 texture_matrix; //light to texture coordinate matrix (transposed)
+	mat2x4 shadow_matrix; //light to shadow coordinate matrix (transposed)
+	vec4 color;
+	vec4 shadow_color;
+	vec2 position;
+	uint flags; //index to light texture
+	float height;
+	float shadow_pixel_size;
+	float pad0;
+	float pad1;
+	float pad2;
+};
+
+layout(set = 2, binding = 3, std140) uniform LightData {
+	Light data[MAX_LIGHTS];
+}
+light_array;
+
+layout(set = 2, binding = 4) uniform texture2D light_textures[MAX_LIGHT_TEXTURES];
+layout(set = 2, binding = 5) uniform texture2D shadow_textures[MAX_LIGHT_TEXTURES];
+
+layout(set = 2, binding = 6) uniform sampler shadow_sampler;
+
+#endif
+
+/* SET3: Render Target Data */
+
+#ifdef SCREEN_TEXTURE_USED
+
+layout(set = 3, binding = 0) uniform texture2D screen_texture;
+
+#endif

servers/rendering/rasterizer_rd/shaders/copy.glsl

@@ -0,0 +1,86 @@
+/* clang-format off */
+[vertex]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(location = 0) out vec2 uv_interp;
+/* clang-format on */
+
+void main() {
+
+	vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0));
+	uv_interp = base_arr[gl_VertexIndex];
+
+	gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0);
+}
+
+/* clang-format off */
+[fragment]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(location = 0) in vec2 uv_interp;
+/* clang-format on */
+
+#ifdef MODE_CUBE_TO_DP
+
+layout(set = 0, binding = 0) uniform samplerCube source_cube;
+
+layout(push_constant, binding = 0, std430) uniform Params {
+	float bias;
+	float z_far;
+	float z_near;
+	bool z_flip;
+}
+params;
+
+layout(location = 0) out float depth_buffer;
+
+#endif
+
+void main() {
+
+#ifdef MODE_CUBE_TO_DP
+
+	vec3 normal = vec3(uv_interp * 2.0 - 1.0, 0.0);
+
+	normal.z = 0.5 - 0.5 * ((normal.x * normal.x) + (normal.y * normal.y));
+	normal = normalize(normal);
+
+	normal.y = -normal.y; //needs to be flipped to match projection matrix
+	if (!params.z_flip) {
+		normal.z = -normal.z;
+	}
+
+	float depth = texture(source_cube, normal).r;
+
+	// absolute values for direction cosines, bigger value equals closer to basis axis
+	vec3 unorm = abs(normal);
+
+	if ((unorm.x >= unorm.y) && (unorm.x >= unorm.z)) {
+		// x code
+		unorm = normal.x > 0.0 ? vec3(1.0, 0.0, 0.0) : vec3(-1.0, 0.0, 0.0);
+	} else if ((unorm.y > unorm.x) && (unorm.y >= unorm.z)) {
+		// y code
+		unorm = normal.y > 0.0 ? vec3(0.0, 1.0, 0.0) : vec3(0.0, -1.0, 0.0);
+	} else if ((unorm.z > unorm.x) && (unorm.z > unorm.y)) {
+		// z code
+		unorm = normal.z > 0.0 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 0.0, -1.0);
+	} else {
+		// oh-no we messed up code
+		// has to be
+		unorm = vec3(1.0, 0.0, 0.0);
+	}
+
+	float depth_fix = 1.0 / dot(normal, unorm);
+
+	depth = 2.0 * depth - 1.0;
+	float linear_depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near));
+	depth_buffer = (linear_depth * depth_fix + params.bias) / params.z_far;
+
+#endif
+}

servers/rendering/rasterizer_rd/shaders/cubemap_downsampler.glsl

@@ -0,0 +1,188 @@
+// Copyright 2016 Activision Publishing, Inc.
+//
+// Permission is hereby granted, free of charge, to any person obtaining
+// a copy of this software and associated documentation files (the "Software"),
+// to deal in the Software without restriction, including without limitation
+// the rights to use, copy, modify, merge, publish, distribute, sublicense,
+// and/or sell copies of the Software, and to permit persons to whom the Software
+// is furnished to do so, subject to the following conditions:
+//
+// The above copyright notice and this permission notice shall be included in all
+// copies or substantial portions of the Software.
+//
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+// SOFTWARE.
+
+/* clang-format off */
+[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+#define BLOCK_SIZE 8
+
+layout(local_size_x = BLOCK_SIZE, local_size_y = BLOCK_SIZE, local_size_z = 1) in;
+/* clang-format on */
+
+layout(set = 0, binding = 0) uniform samplerCube source_cubemap;
+
+layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly imageCube dest_cubemap;
+
+layout(push_constant, binding = 1, std430) uniform Params {
+	uint face_size;
+}
+params;
+
+#define M_PI 3.14159265359
+
+void get_dir_0(out vec3 dir, in float u, in float v) {
+	dir[0] = 1.0;
+	dir[1] = v;
+	dir[2] = -u;
+}
+void get_dir_1(out vec3 dir, in float u, in float v) {
+	dir[0] = -1.0;
+	dir[1] = v;
+	dir[2] = u;
+}
+void get_dir_2(out vec3 dir, in float u, in float v) {
+	dir[0] = u;
+	dir[1] = 1.0;
+	dir[2] = -v;
+}
+void get_dir_3(out vec3 dir, in float u, in float v) {
+	dir[0] = u;
+	dir[1] = -1.0;
+	dir[2] = v;
+}
+void get_dir_4(out vec3 dir, in float u, in float v) {
+	dir[0] = u;
+	dir[1] = v;
+	dir[2] = 1.0;
+}
+void get_dir_5(out vec3 dir, in float u, in float v) {
+	dir[0] = -u;
+	dir[1] = v;
+	dir[2] = -1.0;
+}
+
+float calcWeight(float u, float v) {
+	float val = u * u + v * v + 1.0;
+	return val * sqrt(val);
+}
+
+void main() {
+	uvec3 id = gl_GlobalInvocationID;
+	uint face_size = params.face_size;
+
+	if (id.x < face_size && id.y < face_size) {
+		float inv_face_size = 1.0 / float(face_size);
+
+		float u0 = (float(id.x) * 2.0 + 1.0 - 0.75) * inv_face_size - 1.0;
+		float u1 = (float(id.x) * 2.0 + 1.0 + 0.75) * inv_face_size - 1.0;
+
+		float v0 = (float(id.y) * 2.0 + 1.0 - 0.75) * -inv_face_size + 1.0;
+		float v1 = (float(id.y) * 2.0 + 1.0 + 0.75) * -inv_face_size + 1.0;
+
+		float weights[4];
+		weights[0] = calcWeight(u0, v0);
+		weights[1] = calcWeight(u1, v0);
+		weights[2] = calcWeight(u0, v1);
+		weights[3] = calcWeight(u1, v1);
+
+		const float wsum = 0.5 / (weights[0] + weights[1] + weights[2] + weights[3]);
+		for (int i = 0; i < 4; i++) {
+			weights[i] = weights[i] * wsum + .125;
+		}
+
+		vec3 dir;
+		vec4 color;
+		switch (id.z) {
+			case 0:
+				get_dir_0(dir, u0, v0);
+				color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0];
+
+				get_dir_0(dir, u1, v0);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1];
+
+				get_dir_0(dir, u0, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2];
+
+				get_dir_0(dir, u1, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3];
+				break;
+			case 1:
+				get_dir_1(dir, u0, v0);
+				color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0];
+
+				get_dir_1(dir, u1, v0);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1];
+
+				get_dir_1(dir, u0, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2];
+
+				get_dir_1(dir, u1, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3];
+				break;
+			case 2:
+				get_dir_2(dir, u0, v0);
+				color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0];
+
+				get_dir_2(dir, u1, v0);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1];
+
+				get_dir_2(dir, u0, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2];
+
+				get_dir_2(dir, u1, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3];
+				break;
+			case 3:
+				get_dir_3(dir, u0, v0);
+				color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0];
+
+				get_dir_3(dir, u1, v0);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1];
+
+				get_dir_3(dir, u0, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2];
+
+				get_dir_3(dir, u1, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3];
+				break;
+			case 4:
+				get_dir_4(dir, u0, v0);
+				color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0];
+
+				get_dir_4(dir, u1, v0);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1];
+
+				get_dir_4(dir, u0, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2];
+
+				get_dir_4(dir, u1, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3];
+				break;
+			default:
+				get_dir_5(dir, u0, v0);
+				color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0];
+
+				get_dir_5(dir, u1, v0);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1];
+
+				get_dir_5(dir, u0, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2];
+
+				get_dir_5(dir, u1, v1);
+				color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3];
+				break;
+		}
+		imageStore(dest_cubemap, ivec3(id), color);
+	}
+}

servers/rendering/rasterizer_rd/shaders/cubemap_filter.glsl

@@ -0,0 +1,328 @@
+// Copyright 2016 Activision Publishing, Inc.
+//
+// Permission is hereby granted, free of charge, to any person obtaining
+// a copy of this software and associated documentation files (the "Software"),
+// to deal in the Software without restriction, including without limitation
+// the rights to use, copy, modify, merge, publish, distribute, sublicense,
+// and/or sell copies of the Software, and to permit persons to whom the Software
+// is furnished to do so, subject to the following conditions:
+//
+// The above copyright notice and this permission notice shall be included in all
+// copies or substantial portions of the Software.
+//
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+// SOFTWARE.
+
+/* clang-format off */
+[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+#define GROUP_SIZE 64
+
+layout(local_size_x = GROUP_SIZE, local_size_y = 1, local_size_z = 1) in;
+/* clang-format on */
+
+layout(set = 0, binding = 0) uniform samplerCube source_cubemap;
+layout(rgba16f, set = 2, binding = 0) uniform restrict writeonly imageCube dest_cubemap0;
+layout(rgba16f, set = 2, binding = 1) uniform restrict writeonly imageCube dest_cubemap1;
+layout(rgba16f, set = 2, binding = 2) uniform restrict writeonly imageCube dest_cubemap2;
+layout(rgba16f, set = 2, binding = 3) uniform restrict writeonly imageCube dest_cubemap3;
+layout(rgba16f, set = 2, binding = 4) uniform restrict writeonly imageCube dest_cubemap4;
+layout(rgba16f, set = 2, binding = 5) uniform restrict writeonly imageCube dest_cubemap5;
+layout(rgba16f, set = 2, binding = 6) uniform restrict writeonly imageCube dest_cubemap6;
+
+#ifdef USE_HIGH_QUALITY
+#define NUM_TAPS 32
+#else
+#define NUM_TAPS 8
+#endif
+
+#define BASE_RESOLUTION 128
+
+#ifdef USE_HIGH_QUALITY
+layout(set = 1, binding = 0, std430) buffer restrict readonly Data {
+	vec4[7][5][3][24] coeffs;
+}
+data;
+#else
+layout(set = 1, binding = 0, std430) buffer restrict readonly Data {
+	vec4[7][5][6] coeffs;
+}
+data;
+#endif
+
+void get_dir(out vec3 dir, in vec2 uv, in uint face) {
+	switch (face) {
+		case 0:
+			dir = vec3(1.0, uv[1], -uv[0]);
+			break;
+		case 1:
+			dir = vec3(-1.0, uv[1], uv[0]);
+			break;
+		case 2:
+			dir = vec3(uv[0], 1.0, -uv[1]);
+			break;
+		case 3:
+			dir = vec3(uv[0], -1.0, uv[1]);
+			break;
+		case 4:
+			dir = vec3(uv[0], uv[1], 1.0);
+			break;
+		default:
+			dir = vec3(-uv[0], uv[1], -1.0);
+			break;
+	}
+}
+
+void main() {
+	// INPUT:
+	// id.x = the linear address of the texel (ignoring face)
+	// id.y = the face
+	// -> use to index output texture
+	// id.x = texel x
+	// id.y = texel y
+	// id.z = face
+	uvec3 id = gl_GlobalInvocationID;
+
+	// determine which texel this is
+#ifndef USE_TEXTURE_ARRAY
+	// NOTE (macOS/MoltenVK): Do not rename, "level" variable name conflicts with the Metal "level(float lod)" mipmap sampling function name.
+	int mip_level = 0;
+	if (id.x < (128 * 128)) {
+		mip_level = 0;
+	} else if (id.x < (128 * 128 + 64 * 64)) {
+		mip_level = 1;
+		id.x -= (128 * 128);
+	} else if (id.x < (128 * 128 + 64 * 64 + 32 * 32)) {
+		mip_level = 2;
+		id.x -= (128 * 128 + 64 * 64);
+	} else if (id.x < (128 * 128 + 64 * 64 + 32 * 32 + 16 * 16)) {
+		mip_level = 3;
+		id.x -= (128 * 128 + 64 * 64 + 32 * 32);
+	} else if (id.x < (128 * 128 + 64 * 64 + 32 * 32 + 16 * 16 + 8 * 8)) {
+		mip_level = 4;
+		id.x -= (128 * 128 + 64 * 64 + 32 * 32 + 16 * 16);
+	} else if (id.x < (128 * 128 + 64 * 64 + 32 * 32 + 16 * 16 + 8 * 8 + 4 * 4)) {
+		mip_level = 5;
+		id.x -= (128 * 128 + 64 * 64 + 32 * 32 + 16 * 16 + 8 * 8);
+	} else if (id.x < (128 * 128 + 64 * 64 + 32 * 32 + 16 * 16 + 8 * 8 + 4 * 4 + 2 * 2)) {
+		mip_level = 6;
+		id.x -= (128 * 128 + 64 * 64 + 32 * 32 + 16 * 16 + 8 * 8 + 4 * 4);
+	} else {
+		return;
+	}
+	int res = BASE_RESOLUTION >> mip_level;
+#else // Using Texture Arrays so all levels are the same resolution
+	int res = BASE_RESOLUTION;
+	int mip_level = int(id.x / (BASE_RESOLUTION * BASE_RESOLUTION));
+	id.x -= mip_level * BASE_RESOLUTION * BASE_RESOLUTION;
+#endif
+
+	// determine dir / pos for the texel
+	vec3 dir, adir, frameZ;
+	{
+		id.z = id.y;
+		id.y = id.x / res;
+		id.x -= id.y * res;
+
+		vec2 uv;
+		uv.x = (float(id.x) * 2.0 + 1.0) / float(res) - 1.0;
+		uv.y = -(float(id.y) * 2.0 + 1.0) / float(res) + 1.0;
+
+		get_dir(dir, uv, id.z);
+		frameZ = normalize(dir);
+
+		adir = abs(dir);
+	}
+
+	// GGX gather colors
+	vec4 color = vec4(0.0);
+	for (int axis = 0; axis < 3; axis++) {
+		const int otherAxis0 = 1 - (axis & 1) - (axis >> 1);
+		const int otherAxis1 = 2 - (axis >> 1);
+
+		float frameweight = (max(adir[otherAxis0], adir[otherAxis1]) - .75) / .25;
+		if (frameweight > 0.0) {
+			// determine frame
+			vec3 UpVector;
+			switch (axis) {
+				case 0:
+					UpVector = vec3(1, 0, 0);
+					break;
+				case 1:
+					UpVector = vec3(0, 1, 0);
+					break;
+				default:
+					UpVector = vec3(0, 0, 1);
+					break;
+			}
+
+			vec3 frameX = normalize(cross(UpVector, frameZ));
+			vec3 frameY = cross(frameZ, frameX);
+
+			// calculate parametrization for polynomial
+			float Nx = dir[otherAxis0];
+			float Ny = dir[otherAxis1];
+			float Nz = adir[axis];
+
+			float NmaxXY = max(abs(Ny), abs(Nx));
+			Nx /= NmaxXY;
+			Ny /= NmaxXY;
+
+			float theta;
+			if (Ny < Nx) {
+				if (Ny <= -0.999)
+					theta = Nx;
+				else
+					theta = Ny;
+			} else {
+				if (Ny >= 0.999)
+					theta = -Nx;
+				else
+					theta = -Ny;
+			}
+
+			float phi;
+			if (Nz <= -0.999)
+				phi = -NmaxXY;
+			else if (Nz >= 0.999)
+				phi = NmaxXY;
+			else
+				phi = Nz;
+
+			float theta2 = theta * theta;
+			float phi2 = phi * phi;
+
+			// sample
+			for (int iSuperTap = 0; iSuperTap < NUM_TAPS / 4; iSuperTap++) {
+				const int index = (NUM_TAPS / 4) * axis + iSuperTap;
+
+#ifdef USE_HIGH_QUALITY
+				vec4 coeffsDir0[3];
+				vec4 coeffsDir1[3];
+				vec4 coeffsDir2[3];
+				vec4 coeffsLevel[3];
+				vec4 coeffsWeight[3];
+
+				for (int iCoeff = 0; iCoeff < 3; iCoeff++) {
+					coeffsDir0[iCoeff] = data.coeffs[mip_level][0][iCoeff][index];
+					coeffsDir1[iCoeff] = data.coeffs[mip_level][1][iCoeff][index];
+					coeffsDir2[iCoeff] = data.coeffs[mip_level][2][iCoeff][index];
+					coeffsLevel[iCoeff] = data.coeffs[mip_level][3][iCoeff][index];
+					coeffsWeight[iCoeff] = data.coeffs[mip_level][4][iCoeff][index];
+				}
+
+				for (int iSubTap = 0; iSubTap < 4; iSubTap++) {
+					// determine sample attributes (dir, weight, mip_level)
+					vec3 sample_dir = frameX * (coeffsDir0[0][iSubTap] + coeffsDir0[1][iSubTap] * theta2 + coeffsDir0[2][iSubTap] * phi2) + frameY * (coeffsDir1[0][iSubTap] + coeffsDir1[1][iSubTap] * theta2 + coeffsDir1[2][iSubTap] * phi2) + frameZ * (coeffsDir2[0][iSubTap] + coeffsDir2[1][iSubTap] * theta2 + coeffsDir2[2][iSubTap] * phi2);
+
+					float sample_level = coeffsLevel[0][iSubTap] + coeffsLevel[1][iSubTap] * theta2 + coeffsLevel[2][iSubTap] * phi2;
+
+					float sample_weight = coeffsWeight[0][iSubTap] + coeffsWeight[1][iSubTap] * theta2 + coeffsWeight[2][iSubTap] * phi2;
+#else
+				vec4 coeffsDir0 = data.coeffs[mip_level][0][index];
+				vec4 coeffsDir1 = data.coeffs[mip_level][1][index];
+				vec4 coeffsDir2 = data.coeffs[mip_level][2][index];
+				vec4 coeffsLevel = data.coeffs[mip_level][3][index];
+				vec4 coeffsWeight = data.coeffs[mip_level][4][index];
+
+				for (int iSubTap = 0; iSubTap < 4; iSubTap++) {
+					// determine sample attributes (dir, weight, mip_level)
+					vec3 sample_dir = frameX * coeffsDir0[iSubTap] + frameY * coeffsDir1[iSubTap] + frameZ * coeffsDir2[iSubTap];
+
+					float sample_level = coeffsLevel[iSubTap];
+
+					float sample_weight = coeffsWeight[iSubTap];
+#endif
+
+					sample_weight *= frameweight;
+
+					// adjust for jacobian
+					sample_dir /= max(abs(sample_dir[0]), max(abs(sample_dir[1]), abs(sample_dir[2])));
+					sample_level += 0.75 * log2(dot(sample_dir, sample_dir));
+#ifndef USE_TEXTURE_ARRAY
+					sample_level += float(mip_level) / 6.0; // Hack to increase the perceived roughness and reduce upscaling artifacts
+#endif
+					// sample cubemap
+					color.xyz += textureLod(source_cubemap, normalize(sample_dir), sample_level).xyz * sample_weight;
+					color.w += sample_weight;
+				}
+			}
+		}
+	}
+	color /= color.w;
+
+	// write color
+	color.xyz = max(vec3(0.0), color.xyz);
+	color.w = 1.0;
+#ifdef USE_TEXTURE_ARRAY
+	id.xy *= uvec2(2, 2);
+#endif
+
+	switch (mip_level) {
+		case 0:
+			imageStore(dest_cubemap0, ivec3(id), color);
+#ifdef USE_TEXTURE_ARRAY
+			imageStore(dest_cubemap0, ivec3(id) + ivec3(1.0, 0.0, 0.0), color);
+			imageStore(dest_cubemap0, ivec3(id) + ivec3(0.0, 1.0, 0.0), color);
+			imageStore(dest_cubemap0, ivec3(id) + ivec3(1.0, 1.0, 0.0), color);
+#endif
+			break;
+		case 1:
+			imageStore(dest_cubemap1, ivec3(id), color);
+#ifdef USE_TEXTURE_ARRAY
+			imageStore(dest_cubemap1, ivec3(id) + ivec3(1.0, 0.0, 0.0), color);
+			imageStore(dest_cubemap1, ivec3(id) + ivec3(0.0, 1.0, 0.0), color);
+			imageStore(dest_cubemap1, ivec3(id) + ivec3(1.0, 1.0, 0.0), color);
+#endif
+			break;
+		case 2:
+			imageStore(dest_cubemap2, ivec3(id), color);
+#ifdef USE_TEXTURE_ARRAY
+			imageStore(dest_cubemap2, ivec3(id) + ivec3(1.0, 0.0, 0.0), color);
+			imageStore(dest_cubemap2, ivec3(id) + ivec3(0.0, 1.0, 0.0), color);
+			imageStore(dest_cubemap2, ivec3(id) + ivec3(1.0, 1.0, 0.0), color);
+#endif
+			break;
+		case 3:
+			imageStore(dest_cubemap3, ivec3(id), color);
+#ifdef USE_TEXTURE_ARRAY
+			imageStore(dest_cubemap3, ivec3(id) + ivec3(1.0, 0.0, 0.0), color);
+			imageStore(dest_cubemap3, ivec3(id) + ivec3(0.0, 1.0, 0.0), color);
+			imageStore(dest_cubemap3, ivec3(id) + ivec3(1.0, 1.0, 0.0), color);
+#endif
+			break;
+		case 4:
+			imageStore(dest_cubemap4, ivec3(id), color);
+#ifdef USE_TEXTURE_ARRAY
+			imageStore(dest_cubemap4, ivec3(id) + ivec3(1.0, 0.0, 0.0), color);
+			imageStore(dest_cubemap4, ivec3(id) + ivec3(0.0, 1.0, 0.0), color);
+			imageStore(dest_cubemap4, ivec3(id) + ivec3(1.0, 1.0, 0.0), color);
+#endif
+			break;
+		case 5:
+			imageStore(dest_cubemap5, ivec3(id), color);
+#ifdef USE_TEXTURE_ARRAY
+			imageStore(dest_cubemap5, ivec3(id) + ivec3(1.0, 0.0, 0.0), color);
+			imageStore(dest_cubemap5, ivec3(id) + ivec3(0.0, 1.0, 0.0), color);
+			imageStore(dest_cubemap5, ivec3(id) + ivec3(1.0, 1.0, 0.0), color);
+#endif
+			break;
+		default:
+			imageStore(dest_cubemap6, ivec3(id), color);
+#ifdef USE_TEXTURE_ARRAY
+			imageStore(dest_cubemap6, ivec3(id) + ivec3(1.0, 0.0, 0.0), color);
+			imageStore(dest_cubemap6, ivec3(id) + ivec3(0.0, 1.0, 0.0), color);
+			imageStore(dest_cubemap6, ivec3(id) + ivec3(1.0, 1.0, 0.0), color);
+#endif
+			break;
+	}
+}

servers/rendering/rasterizer_rd/shaders/cubemap_roughness.glsl

@@ -0,0 +1,147 @@
+/* clang-format off */
+[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+#define GROUP_SIZE 8
+
+layout(local_size_x = GROUP_SIZE, local_size_y = GROUP_SIZE, local_size_z = 1) in;
+/* clang-format on */
+
+layout(set = 0, binding = 0) uniform samplerCube source_cube;
+
+layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly imageCube dest_cubemap;
+
+layout(push_constant, binding = 1, std430) uniform Params {
+	uint face_id;
+	uint sample_count;
+	float roughness;
+	bool use_direct_write;
+	float face_size;
+}
+params;
+
+#define M_PI 3.14159265359
+
+vec3 texelCoordToVec(vec2 uv, uint faceID) {
+	mat3 faceUvVectors[6];
+
+	// -x
+	faceUvVectors[1][0] = vec3(0.0, 0.0, 1.0); // u -> +z
+	faceUvVectors[1][1] = vec3(0.0, -1.0, 0.0); // v -> -y
+	faceUvVectors[1][2] = vec3(-1.0, 0.0, 0.0); // -x face
+
+	// +x
+	faceUvVectors[0][0] = vec3(0.0, 0.0, -1.0); // u -> -z
+	faceUvVectors[0][1] = vec3(0.0, -1.0, 0.0); // v -> -y
+	faceUvVectors[0][2] = vec3(1.0, 0.0, 0.0); // +x face
+
+	// -y
+	faceUvVectors[3][0] = vec3(1.0, 0.0, 0.0); // u -> +x
+	faceUvVectors[3][1] = vec3(0.0, 0.0, -1.0); // v -> -z
+	faceUvVectors[3][2] = vec3(0.0, -1.0, 0.0); // -y face
+
+	// +y
+	faceUvVectors[2][0] = vec3(1.0, 0.0, 0.0); // u -> +x
+	faceUvVectors[2][1] = vec3(0.0, 0.0, 1.0); // v -> +z
+	faceUvVectors[2][2] = vec3(0.0, 1.0, 0.0); // +y face
+
+	// -z
+	faceUvVectors[5][0] = vec3(-1.0, 0.0, 0.0); // u -> -x
+	faceUvVectors[5][1] = vec3(0.0, -1.0, 0.0); // v -> -y
+	faceUvVectors[5][2] = vec3(0.0, 0.0, -1.0); // -z face
+
+	// +z
+	faceUvVectors[4][0] = vec3(1.0, 0.0, 0.0); // u -> +x
+	faceUvVectors[4][1] = vec3(0.0, -1.0, 0.0); // v -> -y
+	faceUvVectors[4][2] = vec3(0.0, 0.0, 1.0); // +z face
+
+	// out = u * s_faceUv[0] + v * s_faceUv[1] + s_faceUv[2].
+	vec3 result = (faceUvVectors[faceID][0] * uv.x) + (faceUvVectors[faceID][1] * uv.y) + faceUvVectors[faceID][2];
+	return normalize(result);
+}
+
+vec3 ImportanceSampleGGX(vec2 Xi, float Roughness, vec3 N) {
+	float a = Roughness * Roughness; // DISNEY'S ROUGHNESS [see Burley'12 siggraph]
+
+	// Compute distribution direction
+	float Phi = 2.0 * M_PI * Xi.x;
+	float CosTheta = sqrt((1.0 - Xi.y) / (1.0 + (a * a - 1.0) * Xi.y));
+	float SinTheta = sqrt(1.0 - CosTheta * CosTheta);
+
+	// Convert to spherical direction
+	vec3 H;
+	H.x = SinTheta * cos(Phi);
+	H.y = SinTheta * sin(Phi);
+	H.z = CosTheta;
+
+	vec3 UpVector = abs(N.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(1.0, 0.0, 0.0);
+	vec3 TangentX = normalize(cross(UpVector, N));
+	vec3 TangentY = cross(N, TangentX);
+
+	// Tangent to world space
+	return TangentX * H.x + TangentY * H.y + N * H.z;
+}
+
+// http://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html
+float GGX(float NdotV, float a) {
+	float k = a / 2.0;
+	return NdotV / (NdotV * (1.0 - k) + k);
+}
+
+// http://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html
+float G_Smith(float a, float nDotV, float nDotL) {
+	return GGX(nDotL, a * a) * GGX(nDotV, a * a);
+}
+
+float radicalInverse_VdC(uint bits) {
+	bits = (bits << 16u) | (bits >> 16u);
+	bits = ((bits & 0x55555555u) << 1u) | ((bits & 0xAAAAAAAAu) >> 1u);
+	bits = ((bits & 0x33333333u) << 2u) | ((bits & 0xCCCCCCCCu) >> 2u);
+	bits = ((bits & 0x0F0F0F0Fu) << 4u) | ((bits & 0xF0F0F0F0u) >> 4u);
+	bits = ((bits & 0x00FF00FFu) << 8u) | ((bits & 0xFF00FF00u) >> 8u);
+	return float(bits) * 2.3283064365386963e-10; // / 0x100000000
+}
+
+vec2 Hammersley(uint i, uint N) {
+	return vec2(float(i) / float(N), radicalInverse_VdC(i));
+}
+
+void main() {
+	uvec3 id = gl_GlobalInvocationID;
+	id.z += params.face_id;
+
+	vec2 uv = ((vec2(id.xy) * 2.0 + 1.0) / (params.face_size) - 1.0);
+	vec3 N = texelCoordToVec(uv, id.z);
+
+	//vec4 color = color_interp;
+
+	if (params.use_direct_write) {
+
+		imageStore(dest_cubemap, ivec3(id), vec4(texture(source_cube, N).rgb, 1.0));
+	} else {
+
+		vec4 sum = vec4(0.0, 0.0, 0.0, 0.0);
+
+		for (uint sampleNum = 0u; sampleNum < params.sample_count; sampleNum++) {
+			vec2 xi = Hammersley(sampleNum, params.sample_count);
+
+			vec3 H = ImportanceSampleGGX(xi, params.roughness, N);
+			vec3 V = N;
+			vec3 L = (2.0 * dot(V, H) * H - V);
+
+			float ndotl = clamp(dot(N, L), 0.0, 1.0);
+
+			if (ndotl > 0.0) {
+
+				sum.rgb += textureLod(source_cube, L, 0.0).rgb * ndotl;
+				sum.a += ndotl;
+			}
+		}
+		sum /= sum.a;
+
+		imageStore(dest_cubemap, ivec3(id), vec4(sum.rgb, 1.0));
+	}
+}

servers/rendering/rasterizer_rd/shaders/giprobe.glsl

@@ -0,0 +1,788 @@
+/* clang-format off */
+[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+#ifdef MODE_DYNAMIC
+layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
+#else
+layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in;
+#endif
+/* clang-format on */
+
+#ifndef MODE_DYNAMIC
+
+#define NO_CHILDREN 0xFFFFFFFF
+#define GREY_VEC vec3(0.33333, 0.33333, 0.33333)
+
+struct CellChildren {
+	uint children[8];
+};
+
+layout(set = 0, binding = 1, std430) buffer CellChildrenBuffer {
+	CellChildren data[];
+}
+cell_children;
+
+struct CellData {
+	uint position; // xyz 10 bits
+	uint albedo; //rgb albedo
+	uint emission; //rgb normalized with e as multiplier
+	uint normal; //RGB normal encoded
+};
+
+layout(set = 0, binding = 2, std430) buffer CellDataBuffer {
+	CellData data[];
+}
+cell_data;
+
+#endif // MODE DYNAMIC
+
+#define LIGHT_TYPE_DIRECTIONAL 0
+#define LIGHT_TYPE_OMNI 1
+#define LIGHT_TYPE_SPOT 2
+
+#if defined(MODE_COMPUTE_LIGHT) || defined(MODE_DYNAMIC_LIGHTING)
+
+struct Light {
+
+	uint type;
+	float energy;
+	float radius;
+	float attenuation;
+
+	vec3 color;
+	float spot_angle_radians;
+
+	vec3 position;
+	float spot_attenuation;
+
+	vec3 direction;
+	bool has_shadow;
+};
+
+layout(set = 0, binding = 3, std140) uniform Lights {
+	Light data[MAX_LIGHTS];
+}
+lights;
+
+#endif // MODE COMPUTE LIGHT
+
+#ifdef MODE_SECOND_BOUNCE
+
+layout(set = 0, binding = 5) uniform texture3D color_texture;
+
+#ifdef MODE_ANISOTROPIC
+layout(set = 0, binding = 7) uniform texture3D aniso_pos_texture;
+layout(set = 0, binding = 8) uniform texture3D aniso_neg_texture;
+#endif // MODE ANISOTROPIC
+
+#endif // MODE_SECOND_BOUNCE
+
+#ifndef MODE_DYNAMIC
+
+layout(push_constant, binding = 0, std430) uniform Params {
+	ivec3 limits;
+	uint stack_size;
+
+	float emission_scale;
+	float propagation;
+	float dynamic_range;
+
+	uint light_count;
+	uint cell_offset;
+	uint cell_count;
+	float aniso_strength;
+	uint pad;
+}
+params;
+
+layout(set = 0, binding = 4, std430) buffer Outputs {
+	vec4 data[];
+}
+outputs;
+
+#endif // MODE DYNAMIC
+
+layout(set = 0, binding = 9) uniform texture3D texture_sdf;
+layout(set = 0, binding = 10) uniform sampler texture_sampler;
+
+#ifdef MODE_WRITE_TEXTURE
+
+layout(rgba8, set = 0, binding = 5) uniform restrict writeonly image3D color_tex;
+
+#ifdef MODE_ANISOTROPIC
+
+layout(r16ui, set = 0, binding = 6) uniform restrict writeonly uimage3D aniso_pos_tex;
+layout(r16ui, set = 0, binding = 7) uniform restrict writeonly uimage3D aniso_neg_tex;
+
+#endif
+
+#endif
+
+#ifdef MODE_DYNAMIC
+
+layout(push_constant, binding = 0, std430) uniform Params {
+	ivec3 limits;
+	uint light_count; //when not lighting
+	ivec3 x_dir;
+	float z_base;
+	ivec3 y_dir;
+	float z_sign;
+	ivec3 z_dir;
+	float pos_multiplier;
+	ivec2 rect_pos;
+	ivec2 rect_size;
+	ivec2 prev_rect_ofs;
+	ivec2 prev_rect_size;
+	bool flip_x;
+	bool flip_y;
+	float dynamic_range;
+	bool on_mipmap;
+	float propagation;
+	float pad[3];
+}
+params;
+
+#ifdef MODE_DYNAMIC_LIGHTING
+
+layout(rgba8, set = 0, binding = 5) uniform restrict readonly image2D source_albedo;
+layout(rgba8, set = 0, binding = 6) uniform restrict readonly image2D source_normal;
+layout(rgba8, set = 0, binding = 7) uniform restrict readonly image2D source_orm;
+//layout (set=0,binding=8) uniform texture2D source_depth;
+layout(rgba16f, set = 0, binding = 11) uniform restrict image2D emission;
+layout(r32f, set = 0, binding = 12) uniform restrict image2D depth;
+
+#endif
+
+#ifdef MODE_DYNAMIC_SHRINK
+
+layout(rgba16f, set = 0, binding = 5) uniform restrict readonly image2D source_light;
+layout(r32f, set = 0, binding = 6) uniform restrict readonly image2D source_depth;
+
+#ifdef MODE_DYNAMIC_SHRINK_WRITE
+
+layout(rgba16f, set = 0, binding = 7) uniform restrict writeonly image2D light;
+layout(r32f, set = 0, binding = 8) uniform restrict writeonly image2D depth;
+
+#endif // MODE_DYNAMIC_SHRINK_WRITE
+
+#ifdef MODE_DYNAMIC_SHRINK_PLOT
+
+layout(rgba8, set = 0, binding = 11) uniform restrict image3D color_texture;
+
+#ifdef MODE_ANISOTROPIC
+
+layout(r16ui, set = 0, binding = 12) uniform restrict writeonly uimage3D aniso_pos_texture;
+layout(r16ui, set = 0, binding = 13) uniform restrict writeonly uimage3D aniso_neg_texture;
+
+#endif // MODE ANISOTROPIC
+
+#endif //MODE_DYNAMIC_SHRINK_PLOT
+
+#endif // MODE_DYNAMIC_SHRINK
+
+//layout (rgba8,set=0,binding=5) uniform restrict writeonly image3D color_tex;
+
+#endif // MODE DYNAMIC
+
+#if defined(MODE_COMPUTE_LIGHT) || defined(MODE_DYNAMIC_LIGHTING)
+
+float raymarch(float distance, float distance_adv, vec3 from, vec3 direction) {
+
+	vec3 cell_size = 1.0 / vec3(params.limits);
+	float occlusion = 1.0;
+	while (distance > 0.5) { //use this to avoid precision errors
+		float advance = texture(sampler3D(texture_sdf, texture_sampler), from * cell_size).r * 255.0 - 1.0;
+		if (advance < 0.0) {
+			occlusion = 0.0;
+			break;
+		}
+
+		occlusion = min(advance, occlusion);
+
+		advance = max(distance_adv, advance - mod(advance, distance_adv)); //should always advance in multiples of distance_adv
+
+		from += direction * advance;
+		distance -= advance;
+	}
+
+	return occlusion; //max(0.0,distance);
+}
+
+bool compute_light_vector(uint light, vec3 pos, out float attenuation, out vec3 light_pos) {
+
+	if (lights.data[light].type == LIGHT_TYPE_DIRECTIONAL) {
+
+		light_pos = pos - lights.data[light].direction * length(vec3(params.limits));
+		attenuation = 1.0;
+
+	} else {
+
+		light_pos = lights.data[light].position;
+		float distance = length(pos - light_pos);
+		if (distance >= lights.data[light].radius) {
+			return false;
+		}
+
+		attenuation = pow(clamp(1.0 - distance / lights.data[light].radius, 0.0001, 1.0), lights.data[light].attenuation);
+
+		if (lights.data[light].type == LIGHT_TYPE_SPOT) {
+
+			vec3 rel = normalize(pos - light_pos);
+			float angle = acos(dot(rel, lights.data[light].direction));
+			if (angle > lights.data[light].spot_angle_radians) {
+				return false;
+			}
+
+			float d = clamp(angle / lights.data[light].spot_angle_radians, 0, 1);
+			attenuation *= pow(1.0 - d, lights.data[light].spot_attenuation);
+		}
+	}
+
+	return true;
+}
+
+float get_normal_advance(vec3 p_normal) {
+
+	vec3 normal = p_normal;
+	vec3 unorm = abs(normal);
+
+	if ((unorm.x >= unorm.y) && (unorm.x >= unorm.z)) {
+		// x code
+		unorm = normal.x > 0.0 ? vec3(1.0, 0.0, 0.0) : vec3(-1.0, 0.0, 0.0);
+	} else if ((unorm.y > unorm.x) && (unorm.y >= unorm.z)) {
+		// y code
+		unorm = normal.y > 0.0 ? vec3(0.0, 1.0, 0.0) : vec3(0.0, -1.0, 0.0);
+	} else if ((unorm.z > unorm.x) && (unorm.z > unorm.y)) {
+		// z code
+		unorm = normal.z > 0.0 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 0.0, -1.0);
+	} else {
+		// oh-no we messed up code
+		// has to be
+		unorm = vec3(1.0, 0.0, 0.0);
+	}
+
+	return 1.0 / dot(normal, unorm);
+}
+
+void clip_segment(vec4 plane, vec3 begin, inout vec3 end) {
+
+	vec3 segment = begin - end;
+	float den = dot(plane.xyz, segment);
+
+	//printf("den is %i\n",den);
+	if (den < 0.0001) {
+		return;
+	}
+
+	float dist = (dot(plane.xyz, begin) - plane.w) / den;
+
+	if (dist < 0.0001 || dist > 1.0001) {
+		return;
+	}
+
+	end = begin + segment * -dist;
+}
+
+bool compute_light_at_pos(uint index, vec3 pos, vec3 normal, inout vec3 light, inout vec3 light_dir) {
+	float attenuation;
+	vec3 light_pos;
+
+	if (!compute_light_vector(index, pos, attenuation, light_pos)) {
+		return false;
+	}
+
+	light_dir = normalize(pos - light_pos);
+
+	if (attenuation < 0.01 || (length(normal) > 0.2 && dot(normal, light_dir) >= 0)) {
+		return false; //not facing the light, or attenuation is near zero
+	}
+
+	if (lights.data[index].has_shadow) {
+
+		float distance_adv = get_normal_advance(light_dir);
+
+		vec3 to = pos;
+		if (length(normal) > 0.2) {
+			to += normal * distance_adv * 0.51;
+		} else {
+			to -= sign(light_dir) * 0.45; //go near the edge towards the light direction to avoid self occlusion
+		}
+
+		//clip
+		clip_segment(mix(vec4(-1.0, 0.0, 0.0, 0.0), vec4(1.0, 0.0, 0.0, float(params.limits.x - 1)), bvec4(light_dir.x < 0.0)), to, light_pos);
+		clip_segment(mix(vec4(0.0, -1.0, 0.0, 0.0), vec4(0.0, 1.0, 0.0, float(params.limits.y - 1)), bvec4(light_dir.y < 0.0)), to, light_pos);
+		clip_segment(mix(vec4(0.0, 0.0, -1.0, 0.0), vec4(0.0, 0.0, 1.0, float(params.limits.z - 1)), bvec4(light_dir.z < 0.0)), to, light_pos);
+
+		float distance = length(to - light_pos);
+		if (distance < 0.1) {
+			return false; // hit
+		}
+
+		distance += distance_adv - mod(distance, distance_adv); //make it reach the center of the box always
+		light_pos = to - light_dir * distance;
+
+		//from -= sign(light_dir)*0.45; //go near the edge towards the light direction to avoid self occlusion
+
+		/*float dist = raymarch(distance,distance_adv,light_pos,light_dir);
+
+		if (dist > distance_adv) {
+			return false;
+		}
+
+		attenuation *= 1.0 - smoothstep(0.1*distance_adv,distance_adv,dist);
+		*/
+
+		float occlusion = raymarch(distance, distance_adv, light_pos, light_dir);
+
+		if (occlusion == 0.0) {
+			return false;
+		}
+
+		attenuation *= occlusion; //1.0 - smoothstep(0.1*distance_adv,distance_adv,dist);
+	}
+
+	light = lights.data[index].color * attenuation * lights.data[index].energy;
+	return true;
+}
+
+#endif // MODE COMPUTE LIGHT
+
+void main() {
+
+#ifndef MODE_DYNAMIC
+
+	uint cell_index = gl_GlobalInvocationID.x;
+	if (cell_index >= params.cell_count) {
+		return;
+	}
+	cell_index += params.cell_offset;
+
+	uvec3 posu = uvec3(cell_data.data[cell_index].position & 0x7FF, (cell_data.data[cell_index].position >> 11) & 0x3FF, cell_data.data[cell_index].position >> 21);
+	vec4 albedo = unpackUnorm4x8(cell_data.data[cell_index].albedo);
+
+#endif
+
+	/////////////////COMPUTE LIGHT///////////////////////////////
+
+#ifdef MODE_COMPUTE_LIGHT
+
+	vec3 pos = vec3(posu) + vec3(0.5);
+
+	vec3 emission = vec3(uvec3(cell_data.data[cell_index].emission & 0x1ff, (cell_data.data[cell_index].emission >> 9) & 0x1ff, (cell_data.data[cell_index].emission >> 18) & 0x1ff)) * pow(2.0, float(cell_data.data[cell_index].emission >> 27) - 15.0 - 9.0);
+	vec3 normal = unpackSnorm4x8(cell_data.data[cell_index].normal).xyz;
+
+#ifdef MODE_ANISOTROPIC
+	vec3 accum[6] = vec3[](vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0));
+	const vec3 accum_dirs[6] = vec3[](vec3(1.0, 0.0, 0.0), vec3(-1.0, 0.0, 0.0), vec3(0.0, 1.0, 0.0), vec3(0.0, -1.0, 0.0), vec3(0.0, 0.0, 1.0), vec3(0.0, 0.0, -1.0));
+#else
+	vec3 accum = vec3(0.0);
+#endif
+
+	for (uint i = 0; i < params.light_count; i++) {
+
+		vec3 light;
+		vec3 light_dir;
+		if (!compute_light_at_pos(i, pos, normal.xyz, light, light_dir)) {
+			continue;
+		}
+
+		light *= albedo.rgb;
+
+#ifdef MODE_ANISOTROPIC
+		for (uint j = 0; j < 6; j++) {
+
+			accum[j] += max(0.0, dot(accum_dirs[j], -light_dir)) * light;
+		}
+#else
+		if (length(normal) > 0.2) {
+			accum += max(0.0, dot(normal, -light_dir)) * light;
+		} else {
+			//all directions
+			accum += light;
+		}
+#endif
+	}
+
+#ifdef MODE_ANISOTROPIC
+
+	for (uint i = 0; i < 6; i++) {
+		vec3 light = accum[i];
+		if (length(normal) > 0.2) {
+			light += max(0.0, dot(accum_dirs[i], -normal)) * emission;
+		} else {
+			light += emission;
+		}
+
+		outputs.data[cell_index * 6 + i] = vec4(light, 0.0);
+	}
+
+#else
+	outputs.data[cell_index] = vec4(accum + emission, 0.0);
+
+#endif
+
+#endif //MODE_COMPUTE_LIGHT
+
+	/////////////////SECOND BOUNCE///////////////////////////////
+
+#ifdef MODE_SECOND_BOUNCE
+	vec3 pos = vec3(posu) + vec3(0.5);
+	ivec3 ipos = ivec3(posu);
+	vec4 normal = unpackSnorm4x8(cell_data.data[cell_index].normal);
+
+#ifdef MODE_ANISOTROPIC
+	vec3 accum[6];
+	const vec3 accum_dirs[6] = vec3[](vec3(1.0, 0.0, 0.0), vec3(-1.0, 0.0, 0.0), vec3(0.0, 1.0, 0.0), vec3(0.0, -1.0, 0.0), vec3(0.0, 0.0, 1.0), vec3(0.0, 0.0, -1.0));
+
+	/*vec3 src_color = texelFetch(sampler3D(color_texture,texture_sampler),ipos,0).rgb * params.dynamic_range;
+	vec3 src_aniso_pos = texelFetch(sampler3D(aniso_pos_texture,texture_sampler),ipos,0).rgb;
+	vec3 src_anisp_neg = texelFetch(sampler3D(anisp_neg_texture,texture_sampler),ipos,0).rgb;
+	accum[0]=src_col * src_aniso_pos.x;
+	accum[1]=src_col * src_aniso_neg.x;
+	accum[2]=src_col * src_aniso_pos.y;
+	accum[3]=src_col * src_aniso_neg.y;
+	accum[4]=src_col * src_aniso_pos.z;
+	accum[5]=src_col * src_aniso_neg.z;*/
+
+	accum[0] = outputs.data[cell_index * 6 + 0].rgb;
+	accum[1] = outputs.data[cell_index * 6 + 1].rgb;
+	accum[2] = outputs.data[cell_index * 6 + 2].rgb;
+	accum[3] = outputs.data[cell_index * 6 + 3].rgb;
+	accum[4] = outputs.data[cell_index * 6 + 4].rgb;
+	accum[5] = outputs.data[cell_index * 6 + 5].rgb;
+
+#else
+	vec3 accum = outputs.data[cell_index].rgb;
+
+#endif
+
+	if (length(normal.xyz) > 0.2) {
+
+		vec3 v0 = abs(normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0);
+		vec3 tangent = normalize(cross(v0, normal.xyz));
+		vec3 bitangent = normalize(cross(tangent, normal.xyz));
+		mat3 normal_mat = mat3(tangent, bitangent, normal.xyz);
+
+#define MAX_CONE_DIRS 6
+
+		vec3 cone_dirs[MAX_CONE_DIRS] = vec3[](
+				vec3(0.0, 0.0, 1.0),
+				vec3(0.866025, 0.0, 0.5),
+				vec3(0.267617, 0.823639, 0.5),
+				vec3(-0.700629, 0.509037, 0.5),
+				vec3(-0.700629, -0.509037, 0.5),
+				vec3(0.267617, -0.823639, 0.5));
+
+		float cone_weights[MAX_CONE_DIRS] = float[](0.25, 0.15, 0.15, 0.15, 0.15, 0.15);
+		float tan_half_angle = 0.577;
+
+		for (int i = 0; i < MAX_CONE_DIRS; i++) {
+
+			vec3 direction = normal_mat * cone_dirs[i];
+			vec4 color = vec4(0.0);
+			{
+
+				float dist = 1.5;
+				float max_distance = length(vec3(params.limits));
+				vec3 cell_size = 1.0 / vec3(params.limits);
+
+#ifdef MODE_ANISOTROPIC
+				vec3 aniso_normal = mix(direction, normal.xyz, params.aniso_strength);
+#endif
+				while (dist < max_distance && color.a < 0.95) {
+					float diameter = max(1.0, 2.0 * tan_half_angle * dist);
+					vec3 uvw_pos = (pos + dist * direction) * cell_size;
+					float half_diameter = diameter * 0.5;
+					//check if outside, then break
+					//if ( any(greaterThan(abs(uvw_pos - 0.5),vec3(0.5f + half_diameter * cell_size)) ) ) {
+					//	break;
+					//}
+
+					float log2_diameter = log2(diameter);
+					vec4 scolor = textureLod(sampler3D(color_texture, texture_sampler), uvw_pos, log2_diameter);
+#ifdef MODE_ANISOTROPIC
+
+					vec3 aniso_neg = textureLod(sampler3D(aniso_neg_texture, texture_sampler), uvw_pos, log2_diameter).rgb;
+					vec3 aniso_pos = textureLod(sampler3D(aniso_pos_texture, texture_sampler), uvw_pos, log2_diameter).rgb;
+
+					scolor.rgb *= dot(max(vec3(0.0), (aniso_normal * aniso_pos)), vec3(1.0)) + dot(max(vec3(0.0), (-aniso_normal * aniso_neg)), vec3(1.0));
+#endif
+					float a = (1.0 - color.a);
+					color += a * scolor;
+					dist += half_diameter;
+				}
+			}
+			color *= cone_weights[i] * vec4(albedo.rgb, 1.0) * params.dynamic_range; //restore range
+#ifdef MODE_ANISOTROPIC
+			for (uint j = 0; j < 6; j++) {
+
+				accum[j] += max(0.0, dot(accum_dirs[j], direction)) * color.rgb;
+			}
+#else
+			accum += color.rgb;
+#endif
+		}
+	}
+
+#ifdef MODE_ANISOTROPIC
+
+	outputs.data[cell_index * 6 + 0] = vec4(accum[0], 0.0);
+	outputs.data[cell_index * 6 + 1] = vec4(accum[1], 0.0);
+	outputs.data[cell_index * 6 + 2] = vec4(accum[2], 0.0);
+	outputs.data[cell_index * 6 + 3] = vec4(accum[3], 0.0);
+	outputs.data[cell_index * 6 + 4] = vec4(accum[4], 0.0);
+	outputs.data[cell_index * 6 + 5] = vec4(accum[5], 0.0);
+#else
+	outputs.data[cell_index] = vec4(accum, 0.0);
+
+#endif
+
+#endif // MODE_SECOND_BOUNCE
+
+	/////////////////UPDATE MIPMAPS///////////////////////////////
+
+#ifdef MODE_UPDATE_MIPMAPS
+
+	{
+#ifdef MODE_ANISOTROPIC
+		vec3 light_accum[6] = vec3[](vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0));
+#else
+		vec3 light_accum = vec3(0.0);
+#endif
+		float count = 0.0;
+		for (uint i = 0; i < 8; i++) {
+			uint child_index = cell_children.data[cell_index].children[i];
+			if (child_index == NO_CHILDREN) {
+				continue;
+			}
+#ifdef MODE_ANISOTROPIC
+			light_accum[0] += outputs.data[child_index * 6 + 0].rgb;
+			light_accum[1] += outputs.data[child_index * 6 + 1].rgb;
+			light_accum[2] += outputs.data[child_index * 6 + 2].rgb;
+			light_accum[3] += outputs.data[child_index * 6 + 3].rgb;
+			light_accum[4] += outputs.data[child_index * 6 + 4].rgb;
+			light_accum[5] += outputs.data[child_index * 6 + 5].rgb;
+
+#else
+			light_accum += outputs.data[child_index].rgb;
+
+#endif
+
+			count += 1.0;
+		}
+
+		float divisor = mix(8.0, count, params.propagation);
+#ifdef MODE_ANISOTROPIC
+		outputs.data[cell_index * 6 + 0] = vec4(light_accum[0] / divisor, 0.0);
+		outputs.data[cell_index * 6 + 1] = vec4(light_accum[1] / divisor, 0.0);
+		outputs.data[cell_index * 6 + 2] = vec4(light_accum[2] / divisor, 0.0);
+		outputs.data[cell_index * 6 + 3] = vec4(light_accum[3] / divisor, 0.0);
+		outputs.data[cell_index * 6 + 4] = vec4(light_accum[4] / divisor, 0.0);
+		outputs.data[cell_index * 6 + 5] = vec4(light_accum[5] / divisor, 0.0);
+
+#else
+		outputs.data[cell_index] = vec4(light_accum / divisor, 0.0);
+#endif
+	}
+#endif
+
+	///////////////////WRITE TEXTURE/////////////////////////////
+
+#ifdef MODE_WRITE_TEXTURE
+	{
+
+#ifdef MODE_ANISOTROPIC
+		vec3 accum_total = vec3(0.0);
+		accum_total += outputs.data[cell_index * 6 + 0].rgb;
+		accum_total += outputs.data[cell_index * 6 + 1].rgb;
+		accum_total += outputs.data[cell_index * 6 + 2].rgb;
+		accum_total += outputs.data[cell_index * 6 + 3].rgb;
+		accum_total += outputs.data[cell_index * 6 + 4].rgb;
+		accum_total += outputs.data[cell_index * 6 + 5].rgb;
+
+		float accum_total_energy = max(dot(accum_total, GREY_VEC), 0.00001);
+		vec3 iso_positive = vec3(dot(outputs.data[cell_index * 6 + 0].rgb, GREY_VEC), dot(outputs.data[cell_index * 6 + 2].rgb, GREY_VEC), dot(outputs.data[cell_index * 6 + 4].rgb, GREY_VEC)) / vec3(accum_total_energy);
+		vec3 iso_negative = vec3(dot(outputs.data[cell_index * 6 + 1].rgb, GREY_VEC), dot(outputs.data[cell_index * 6 + 3].rgb, GREY_VEC), dot(outputs.data[cell_index * 6 + 5].rgb, GREY_VEC)) / vec3(accum_total_energy);
+
+		{
+			uint aniso_pos = uint(clamp(iso_positive.b * 31.0, 0.0, 31.0));
+			aniso_pos |= uint(clamp(iso_positive.g * 63.0, 0.0, 63.0)) << 5;
+			aniso_pos |= uint(clamp(iso_positive.r * 31.0, 0.0, 31.0)) << 11;
+			imageStore(aniso_pos_tex, ivec3(posu), uvec4(aniso_pos));
+		}
+
+		{
+			uint aniso_neg = uint(clamp(iso_negative.b * 31.0, 0.0, 31.0));
+			aniso_neg |= uint(clamp(iso_negative.g * 63.0, 0.0, 63.0)) << 5;
+			aniso_neg |= uint(clamp(iso_negative.r * 31.0, 0.0, 31.0)) << 11;
+			imageStore(aniso_neg_tex, ivec3(posu), uvec4(aniso_neg));
+		}
+
+		imageStore(color_tex, ivec3(posu), vec4(accum_total / params.dynamic_range, albedo.a));
+
+#else
+
+		imageStore(color_tex, ivec3(posu), vec4(outputs.data[cell_index].rgb / params.dynamic_range, albedo.a));
+
+#endif
+	}
+#endif
+
+	///////////////////DYNAMIC LIGHTING/////////////////////////////
+
+#ifdef MODE_DYNAMIC
+
+	ivec2 pos_xy = ivec2(gl_GlobalInvocationID.xy);
+	if (any(greaterThanEqual(pos_xy, params.rect_size))) {
+		return; //out of bounds
+	}
+
+	ivec2 uv_xy = pos_xy;
+	if (params.flip_x) {
+		uv_xy.x = params.rect_size.x - pos_xy.x - 1;
+	}
+	if (params.flip_y) {
+		uv_xy.y = params.rect_size.y - pos_xy.y - 1;
+	}
+
+#ifdef MODE_DYNAMIC_LIGHTING
+
+	{
+		float z = params.z_base + imageLoad(depth, uv_xy).x * params.z_sign;
+
+		ivec3 pos = params.x_dir * (params.rect_pos.x + pos_xy.x) + params.y_dir * (params.rect_pos.y + pos_xy.y) + abs(params.z_dir) * int(z);
+
+		vec3 normal = imageLoad(source_normal, uv_xy).xyz * 2.0 - 1.0;
+		normal = vec3(params.x_dir) * normal.x * mix(1.0, -1.0, params.flip_x) + vec3(params.y_dir) * normal.y * mix(1.0, -1.0, params.flip_y) - vec3(params.z_dir) * normal.z;
+
+		vec4 albedo = imageLoad(source_albedo, uv_xy);
+
+		//determine the position in space
+
+		vec3 accum = vec3(0.0);
+		for (uint i = 0; i < params.light_count; i++) {
+
+			vec3 light;
+			vec3 light_dir;
+			if (!compute_light_at_pos(i, vec3(pos) * params.pos_multiplier, normal, light, light_dir)) {
+				continue;
+			}
+
+			light *= albedo.rgb;
+
+			accum += max(0.0, dot(normal, -light_dir)) * light;
+		}
+
+		accum += imageLoad(emission, uv_xy).xyz;
+
+		imageStore(emission, uv_xy, vec4(accum, albedo.a));
+		imageStore(depth, uv_xy, vec4(z));
+	}
+
+#endif // MODE DYNAMIC LIGHTING
+
+#ifdef MODE_DYNAMIC_SHRINK
+
+	{
+		vec4 accum = vec4(0.0);
+		float accum_z = 0.0;
+		float count = 0.0;
+
+		for (int i = 0; i < 4; i++) {
+			ivec2 ofs = pos_xy * 2 + ivec2(i & 1, i >> 1) - params.prev_rect_ofs;
+			if (any(lessThan(ofs, ivec2(0))) || any(greaterThanEqual(ofs, params.prev_rect_size))) {
+				continue;
+			}
+			if (params.flip_x) {
+				ofs.x = params.prev_rect_size.x - ofs.x - 1;
+			}
+			if (params.flip_y) {
+				ofs.y = params.prev_rect_size.y - ofs.y - 1;
+			}
+
+			vec4 light = imageLoad(source_light, ofs);
+			if (light.a == 0.0) { //ignore empty
+				continue;
+			}
+			accum += light;
+			float z = imageLoad(source_depth, ofs).x;
+			accum_z += z * 0.5; //shrink half too
+			count += 1.0;
+		}
+
+		if (params.on_mipmap) {
+			accum.rgb /= mix(8.0, count, params.propagation);
+			accum.a /= 8.0;
+		} else {
+			accum /= 4.0;
+		}
+
+		if (count == 0.0) {
+			accum_z = 0.0; //avoid nan
+		} else {
+			accum_z /= count;
+		}
+
+#ifdef MODE_DYNAMIC_SHRINK_WRITE
+
+		imageStore(light, uv_xy, accum);
+		imageStore(depth, uv_xy, vec4(accum_z));
+#endif
+
+#ifdef MODE_DYNAMIC_SHRINK_PLOT
+
+		if (accum.a < 0.001) {
+			return; //do not blit if alpha is too low
+		}
+
+		ivec3 pos = params.x_dir * (params.rect_pos.x + pos_xy.x) + params.y_dir * (params.rect_pos.y + pos_xy.y) + abs(params.z_dir) * int(accum_z);
+
+		float z_frac = fract(accum_z);
+
+		for (int i = 0; i < 2; i++) {
+			ivec3 pos3d = pos + abs(params.z_dir) * i;
+			if (any(lessThan(pos3d, ivec3(0))) || any(greaterThanEqual(pos3d, params.limits))) {
+				//skip if offlimits
+				continue;
+			}
+			vec4 color_blit = accum * (i == 0 ? 1.0 - z_frac : z_frac);
+			vec4 color = imageLoad(color_texture, pos3d);
+			color.rgb *= params.dynamic_range;
+
+#if 0
+			color.rgb = mix(color.rgb,color_blit.rgb,color_blit.a);
+			color.a+=color_blit.a;
+#else
+
+			float sa = 1.0 - color_blit.a;
+			vec4 result;
+			result.a = color.a * sa + color_blit.a;
+			if (result.a == 0.0) {
+				result = vec4(0.0);
+			} else {
+				result.rgb = (color.rgb * color.a * sa + color_blit.rgb * color_blit.a) / result.a;
+				color = result;
+			}
+
+#endif
+			color.rgb /= params.dynamic_range;
+			imageStore(color_texture, pos3d, color);
+			//imageStore(color_texture,pos3d,vec4(1,1,1,1));
+
+#ifdef MODE_ANISOTROPIC
+			//do not care about anisotropy for dynamic objects, just store full lit in all directions
+			imageStore(aniso_pos_texture, pos3d, uvec4(0xFFFF));
+			imageStore(aniso_neg_texture, pos3d, uvec4(0xFFFF));
+
+#endif // ANISOTROPIC
+		}
+#endif // MODE_DYNAMIC_SHRINK_PLOT
+	}
+#endif
+
+#endif // MODE DYNAMIC
+}

servers/rendering/rasterizer_rd/shaders/giprobe_debug.glsl

@@ -0,0 +1,208 @@
+/* clang-format off */
+[vertex]
+
+#version 450
+
+VERSION_DEFINES
+
+struct CellData {
+	uint position; // xyz 10 bits
+	uint albedo; //rgb albedo
+	uint emission; //rgb normalized with e as multiplier
+	uint normal; //RGB normal encoded
+};
+/* clang-format on */
+
+layout(set = 0, binding = 1, std140) buffer CellDataBuffer {
+	CellData data[];
+}
+cell_data;
+
+layout(set = 0, binding = 2) uniform texture3D color_tex;
+
+layout(set = 0, binding = 3) uniform sampler tex_sampler;
+
+#ifdef USE_ANISOTROPY
+layout(set = 0, binding = 4) uniform texture3D aniso_pos_tex;
+layout(set = 0, binding = 5) uniform texture3D aniso_neg_tex;
+#endif
+
+layout(push_constant, binding = 0, std430) uniform Params {
+
+	mat4 projection;
+	uint cell_offset;
+	float dynamic_range;
+	float alpha;
+	uint level;
+	ivec3 bounds;
+	uint pad;
+}
+params;
+
+layout(location = 0) out vec4 color_interp;
+
+void main() {
+
+	const vec3 cube_triangles[36] = vec3[](
+			vec3(-1.0f, -1.0f, -1.0f),
+			vec3(-1.0f, -1.0f, 1.0f),
+			vec3(-1.0f, 1.0f, 1.0f),
+			vec3(1.0f, 1.0f, -1.0f),
+			vec3(-1.0f, -1.0f, -1.0f),
+			vec3(-1.0f, 1.0f, -1.0f),
+			vec3(1.0f, -1.0f, 1.0f),
+			vec3(-1.0f, -1.0f, -1.0f),
+			vec3(1.0f, -1.0f, -1.0f),
+			vec3(1.0f, 1.0f, -1.0f),
+			vec3(1.0f, -1.0f, -1.0f),
+			vec3(-1.0f, -1.0f, -1.0f),
+			vec3(-1.0f, -1.0f, -1.0f),
+			vec3(-1.0f, 1.0f, 1.0f),
+			vec3(-1.0f, 1.0f, -1.0f),
+			vec3(1.0f, -1.0f, 1.0f),
+			vec3(-1.0f, -1.0f, 1.0f),
+			vec3(-1.0f, -1.0f, -1.0f),
+			vec3(-1.0f, 1.0f, 1.0f),
+			vec3(-1.0f, -1.0f, 1.0f),
+			vec3(1.0f, -1.0f, 1.0f),
+			vec3(1.0f, 1.0f, 1.0f),
+			vec3(1.0f, -1.0f, -1.0f),
+			vec3(1.0f, 1.0f, -1.0f),
+			vec3(1.0f, -1.0f, -1.0f),
+			vec3(1.0f, 1.0f, 1.0f),
+			vec3(1.0f, -1.0f, 1.0f),
+			vec3(1.0f, 1.0f, 1.0f),
+			vec3(1.0f, 1.0f, -1.0f),
+			vec3(-1.0f, 1.0f, -1.0f),
+			vec3(1.0f, 1.0f, 1.0f),
+			vec3(-1.0f, 1.0f, -1.0f),
+			vec3(-1.0f, 1.0f, 1.0f),
+			vec3(1.0f, 1.0f, 1.0f),
+			vec3(-1.0f, 1.0f, 1.0f),
+			vec3(1.0f, -1.0f, 1.0f));
+
+	vec3 vertex = cube_triangles[gl_VertexIndex] * 0.5 + 0.5;
+#ifdef MODE_DEBUG_LIGHT_FULL
+	uvec3 posu = uvec3(gl_InstanceIndex % params.bounds.x, (gl_InstanceIndex / params.bounds.x) % params.bounds.y, gl_InstanceIndex / (params.bounds.y * params.bounds.x));
+#else
+	uint cell_index = gl_InstanceIndex + params.cell_offset;
+
+	uvec3 posu = uvec3(cell_data.data[cell_index].position & 0x7FF, (cell_data.data[cell_index].position >> 11) & 0x3FF, cell_data.data[cell_index].position >> 21);
+#endif
+
+#ifdef MODE_DEBUG_EMISSION
+	color_interp.xyz = vec3(uvec3(cell_data.data[cell_index].emission & 0x1ff, (cell_data.data[cell_index].emission >> 9) & 0x1ff, (cell_data.data[cell_index].emission >> 18) & 0x1ff)) * pow(2.0, float(cell_data.data[cell_index].emission >> 27) - 15.0 - 9.0);
+#endif
+
+#ifdef MODE_DEBUG_COLOR
+	color_interp.xyz = unpackUnorm4x8(cell_data.data[cell_index].albedo).xyz;
+#endif
+
+#ifdef MODE_DEBUG_LIGHT
+
+#ifdef USE_ANISOTROPY
+
+#define POS_X 0
+#define POS_Y 1
+#define POS_Z 2
+#define NEG_X 3
+#define NEG_Y 4
+#define NEG_Z 5
+
+	const uint triangle_aniso[12] = uint[](
+			NEG_X,
+			NEG_Z,
+			NEG_Y,
+			NEG_Z,
+			NEG_X,
+			NEG_Y,
+			POS_Z,
+			POS_X,
+			POS_X,
+			POS_Y,
+			POS_Y,
+			POS_Z);
+
+	color_interp.xyz = texelFetch(sampler3D(color_tex, tex_sampler), ivec3(posu), int(params.level)).xyz * params.dynamic_range;
+	vec3 aniso_pos = texelFetch(sampler3D(aniso_pos_tex, tex_sampler), ivec3(posu), int(params.level)).xyz;
+	vec3 aniso_neg = texelFetch(sampler3D(aniso_neg_tex, tex_sampler), ivec3(posu), int(params.level)).xyz;
+	uint side = triangle_aniso[gl_VertexIndex / 3];
+
+	float strength = 0.0;
+	switch (side) {
+		case POS_X: strength = aniso_pos.x; break;
+		case POS_Y: strength = aniso_pos.y; break;
+		case POS_Z: strength = aniso_pos.z; break;
+		case NEG_X: strength = aniso_neg.x; break;
+		case NEG_Y: strength = aniso_neg.y; break;
+		case NEG_Z: strength = aniso_neg.z; break;
+	}
+
+	color_interp.xyz *= strength;
+
+#else
+	color_interp = texelFetch(sampler3D(color_tex, tex_sampler), ivec3(posu), int(params.level));
+	color_interp.xyz *params.dynamic_range;
+
+#endif
+
+#endif
+	float scale = (1 << params.level);
+
+	gl_Position = params.projection * vec4((vec3(posu) + vertex) * scale, 1.0);
+
+#ifdef MODE_DEBUG_LIGHT_FULL
+	if (color_interp.a == 0.0) {
+		gl_Position = vec4(0.0); //force clip and not draw
+	}
+#else
+	color_interp.a = params.alpha;
+#endif
+}
+
+/* clang-format off */
+[fragment]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(location = 0) in vec4 color_interp;
+/* clang-format on */
+layout(location = 0) out vec4 frag_color;
+
+void main() {
+
+	frag_color = color_interp;
+
+#ifdef MODE_DEBUG_LIGHT_FULL
+
+	//there really is no alpha, so use dither
+
+	int x = int(gl_FragCoord.x) % 4;
+	int y = int(gl_FragCoord.y) % 4;
+	int index = x + y * 4;
+	float limit = 0.0;
+	if (x < 8) {
+		if (index == 0) limit = 0.0625;
+		if (index == 1) limit = 0.5625;
+		if (index == 2) limit = 0.1875;
+		if (index == 3) limit = 0.6875;
+		if (index == 4) limit = 0.8125;
+		if (index == 5) limit = 0.3125;
+		if (index == 6) limit = 0.9375;
+		if (index == 7) limit = 0.4375;
+		if (index == 8) limit = 0.25;
+		if (index == 9) limit = 0.75;
+		if (index == 10) limit = 0.125;
+		if (index == 11) limit = 0.625;
+		if (index == 12) limit = 1.0;
+		if (index == 13) limit = 0.5;
+		if (index == 14) limit = 0.875;
+		if (index == 15) limit = 0.375;
+	}
+	if (frag_color.a < limit) {
+		discard;
+	}
+#endif
+}

servers/rendering/rasterizer_rd/shaders/giprobe_sdf.glsl

@@ -0,0 +1,187 @@
+/* clang-format off */
+[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(local_size_x = 4, local_size_y = 4, local_size_z = 4) in;
+/* clang-format on */
+
+#define MAX_DISTANCE 100000
+
+#define NO_CHILDREN 0xFFFFFFFF
+#define GREY_VEC vec3(0.33333, 0.33333, 0.33333)
+
+struct CellChildren {
+	uint children[8];
+};
+
+layout(set = 0, binding = 1, std430) buffer CellChildrenBuffer {
+	CellChildren data[];
+}
+cell_children;
+
+struct CellData {
+	uint position; // xyz 10 bits
+	uint albedo; //rgb albedo
+	uint emission; //rgb normalized with e as multiplier
+	uint normal; //RGB normal encoded
+};
+
+layout(set = 0, binding = 2, std430) buffer CellDataBuffer {
+	CellData data[];
+}
+cell_data;
+
+layout(r8ui, set = 0, binding = 3) uniform restrict writeonly uimage3D sdf_tex;
+
+layout(push_constant, binding = 0, std430) uniform Params {
+	uint offset;
+	uint end;
+	uint pad0;
+	uint pad1;
+}
+params;
+
+void main() {
+
+	vec3 pos = vec3(gl_GlobalInvocationID);
+	float closest_dist = 100000.0;
+
+	for (uint i = params.offset; i < params.end; i++) {
+		vec3 posu = vec3(uvec3(cell_data.data[i].position & 0x7FF, (cell_data.data[i].position >> 11) & 0x3FF, cell_data.data[i].position >> 21));
+		float dist = length(pos - posu);
+		if (dist < closest_dist) {
+			closest_dist = dist;
+		}
+	}
+
+	uint dist_8;
+
+	if (closest_dist < 0.0001) { // same cell
+		dist_8 = 0; //equals to -1
+	} else {
+		dist_8 = clamp(uint(closest_dist), 0, 254) + 1; //conservative, 0 is 1, so <1 is considered solid
+	}
+
+	imageStore(sdf_tex, ivec3(gl_GlobalInvocationID), uvec4(dist_8));
+	//imageStore(sdf_tex,pos,uvec4(pos*2,0));
+}
+
+#if 0
+layout(push_constant, binding = 0, std430) uniform Params {
+
+	ivec3 limits;
+	uint stack_size;
+} params;
+
+float distance_to_aabb(ivec3 pos, ivec3 aabb_pos, ivec3 aabb_size) {
+
+	vec3 delta = vec3(max(ivec3(0), max(aabb_pos - pos, pos - (aabb_pos + aabb_size - ivec3(1)))));
+	return length(delta);
+}
+
+void main() {
+
+	ivec3 pos = ivec3(gl_GlobalInvocationID);
+
+	uint stack[10] = uint[](0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
+	uint stack_indices[10] = uint[](0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
+	ivec3 stack_positions[10] = ivec3[](ivec3(0), ivec3(0), ivec3(0), ivec3(0), ivec3(0), ivec3(0), ivec3(0), ivec3(0), ivec3(0), ivec3(0));
+
+	const uint cell_orders[8] = uint[](
+			0x11f58d1,
+			0xe2e70a,
+			0xd47463,
+			0xbb829c,
+			0x8d11f5,
+			0x70ae2e,
+			0x463d47,
+			0x29cbb8);
+
+	bool cell_found = false;
+	bool cell_found_exact = false;
+	ivec3 closest_cell_pos;
+	float closest_distance = MAX_DISTANCE;
+	int stack_pos = 0;
+
+	while (true) {
+
+		uint index = stack_indices[stack_pos] >> 24;
+
+		if (index == 8) {
+			//go up
+			if (stack_pos == 0) {
+				break; //done going through octree
+			}
+			stack_pos--;
+			continue;
+		}
+
+		stack_indices[stack_pos] = (stack_indices[stack_pos] & ((1 << 24) - 1)) | ((index + 1) << 24);
+
+		uint cell_index = (stack_indices[stack_pos] >> (index * 3)) & 0x7;
+		uint child_cell = cell_children.data[stack[stack_pos]].children[cell_index];
+
+		if (child_cell == NO_CHILDREN) {
+			continue;
+		}
+
+		ivec3 child_cell_size = params.limits >> (stack_pos + 1);
+		ivec3 child_cell_pos = stack_positions[stack_pos];
+
+		child_cell_pos += mix(ivec3(0), child_cell_size, bvec3(uvec3(index & 1, index & 2, index & 4) != uvec3(0)));
+
+		bool is_leaf = stack_pos == (params.stack_size - 2);
+
+		if (child_cell_pos == pos && is_leaf) {
+			//we may actually end up in the exact cell.
+			//if this happens, just abort
+			cell_found_exact = true;
+			break;
+		}
+
+		if (cell_found) {
+			//discard by distance
+			float distance = distance_to_aabb(pos, child_cell_pos, child_cell_size);
+			if (distance >= closest_distance) {
+				continue; //pointless, just test next child
+			} else if (is_leaf) {
+				//closer than what we have AND end of stack, save and continue
+				closest_cell_pos = child_cell_pos;
+				closest_distance = distance;
+				continue;
+			}
+		} else if (is_leaf) {
+			//first solid cell we find, save and continue
+			closest_distance = distance_to_aabb(pos, child_cell_pos, child_cell_size);
+			closest_cell_pos = child_cell_pos;
+			cell_found = true;
+			continue;
+		}
+
+		bvec3 direction = greaterThan((pos - (child_cell_pos + (child_cell_size >> 1))), ivec3(0));
+		uint cell_order = 0;
+		cell_order |= mix(0, 1, direction.x);
+		cell_order |= mix(0, 2, direction.y);
+		cell_order |= mix(0, 4, direction.z);
+
+		stack[stack_pos + 1] = child_cell;
+		stack_indices[stack_pos + 1] = cell_orders[cell_order]; //start counting
+		stack_positions[stack_pos + 1] = child_cell_pos;
+		stack_pos++; //go up stack
+	}
+
+	uint dist_8;
+
+	if (cell_found_exact) {
+		dist_8 = 0; //equals to -1
+	} else {
+		float closest_distance = length(vec3(pos - closest_cell_pos));
+		dist_8 = clamp(uint(closest_distance), 0, 254) + 1; //conservative, 0 is 1, so <1 is considered solid
+	}
+
+	imageStore(sdf_tex, pos, uvec4(dist_8));
+}
+#endif

servers/rendering/rasterizer_rd/shaders/giprobe_write.glsl

@@ -0,0 +1,335 @@
+/* clang-format off */
+[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in;
+/* clang-format on */
+
+#define NO_CHILDREN 0xFFFFFFFF
+#define GREY_VEC vec3(0.33333, 0.33333, 0.33333)
+
+struct CellChildren {
+	uint children[8];
+};
+
+layout(set = 0, binding = 1, std430) buffer CellChildrenBuffer {
+	CellChildren data[];
+}
+cell_children;
+
+struct CellData {
+	uint position; // xyz 10 bits
+	uint albedo; //rgb albedo
+	uint emission; //rgb normalized with e as multiplier
+	uint normal; //RGB normal encoded
+};
+
+layout(set = 0, binding = 2, std430) buffer CellDataBuffer {
+	CellData data[];
+}
+cell_data;
+
+#define LIGHT_TYPE_DIRECTIONAL 0
+#define LIGHT_TYPE_OMNI 1
+#define LIGHT_TYPE_SPOT 2
+
+#ifdef MODE_COMPUTE_LIGHT
+
+struct Light {
+	uint type;
+	float energy;
+	float radius;
+	float attenuation;
+
+	vec3 color;
+	float spot_angle_radians;
+
+	vec3 position;
+	float spot_attenuation;
+
+	vec3 direction;
+	bool has_shadow;
+};
+
+layout(set = 0, binding = 3, std140) uniform Lights {
+	Light data[MAX_LIGHTS];
+}
+lights;
+
+#endif
+
+layout(push_constant, binding = 0, std430) uniform Params {
+	ivec3 limits;
+	uint stack_size;
+
+	float emission_scale;
+	float propagation;
+	float dynamic_range;
+
+	uint light_count;
+	uint cell_offset;
+	uint cell_count;
+	uint pad[2];
+}
+params;
+
+layout(set = 0, binding = 4, std140) uniform Outputs {
+	vec4 data[];
+}
+output;
+
+#ifdef MODE_COMPUTE_LIGHT
+
+uint raymarch(float distance, float distance_adv, vec3 from, vec3 direction) {
+
+	uint result = NO_CHILDREN;
+
+	ivec3 size = ivec3(max(max(params.limits.x, params.limits.y), params.limits.z));
+
+	while (distance > -distance_adv) { //use this to avoid precision errors
+
+		uint cell = 0;
+
+		ivec3 pos = ivec3(from);
+
+		if (all(greaterThanEqual(pos, ivec3(0))) && all(lessThan(pos, size))) {
+
+			ivec3 ofs = ivec3(0);
+			ivec3 half_size = size / 2;
+
+			for (int i = 0; i < params.stack_size - 1; i++) {
+
+				bvec3 greater = greaterThanEqual(pos, ofs + half_size);
+
+				ofs += mix(ivec3(0), half_size, greater);
+
+				uint child = 0; //wonder if this can be done faster
+				if (greater.x) {
+					child |= 1;
+				}
+				if (greater.y) {
+					child |= 2;
+				}
+				if (greater.z) {
+					child |= 4;
+				}
+
+				cell = cell_children.data[cell].children[child];
+				if (cell == NO_CHILDREN)
+					break;
+
+				half_size >>= ivec3(1);
+			}
+
+			if (cell != NO_CHILDREN) {
+				return cell; //found cell!
+			}
+		}
+
+		from += direction * distance_adv;
+		distance -= distance_adv;
+	}
+
+	return NO_CHILDREN;
+}
+
+bool compute_light_vector(uint light, uint cell, vec3 pos, out float attenuation, out vec3 light_pos) {
+
+	if (lights.data[light].type == LIGHT_TYPE_DIRECTIONAL) {
+
+		light_pos = pos - lights.data[light].direction * length(vec3(params.limits));
+		attenuation = 1.0;
+
+	} else {
+
+		light_pos = lights.data[light].position;
+		float distance = length(pos - light_pos);
+		if (distance >= lights.data[light].radius) {
+			return false;
+		}
+
+		attenuation = pow(clamp(1.0 - distance / lights.data[light].radius, 0.0001, 1.0), lights.data[light].attenuation);
+
+		if (lights.data[light].type == LIGHT_TYPE_SPOT) {
+
+			vec3 rel = normalize(pos - light_pos);
+			float angle = acos(dot(rel, lights.data[light].direction));
+			if (angle > lights.data[light].spot_angle_radians) {
+				return false;
+			}
+
+			float d = clamp(angle / lights.data[light].spot_angle_radians, 0, 1);
+			attenuation *= pow(1.0 - d, lights.data[light].spot_attenuation);
+		}
+	}
+
+	return true;
+}
+
+float get_normal_advance(vec3 p_normal) {
+
+	vec3 normal = p_normal;
+	vec3 unorm = abs(normal);
+
+	if ((unorm.x >= unorm.y) && (unorm.x >= unorm.z)) {
+		// x code
+		unorm = normal.x > 0.0 ? vec3(1.0, 0.0, 0.0) : vec3(-1.0, 0.0, 0.0);
+	} else if ((unorm.y > unorm.x) && (unorm.y >= unorm.z)) {
+		// y code
+		unorm = normal.y > 0.0 ? vec3(0.0, 1.0, 0.0) : vec3(0.0, -1.0, 0.0);
+	} else if ((unorm.z > unorm.x) && (unorm.z > unorm.y)) {
+		// z code
+		unorm = normal.z > 0.0 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 0.0, -1.0);
+	} else {
+		// oh-no we messed up code
+		// has to be
+		unorm = vec3(1.0, 0.0, 0.0);
+	}
+
+	return 1.0 / dot(normal, unorm);
+}
+
+#endif
+
+void main() {
+
+	uint cell_index = gl_GlobalInvocationID.x;
+	if (cell_index >= params.cell_count) {
+		return;
+	}
+	cell_index += params.cell_offset;
+
+	uvec3 posu = uvec3(cell_data.data[cell_index].position & 0x7FF, (cell_data.data[cell_index].position >> 11) & 0x3FF, cell_data.data[cell_index].position >> 21);
+	vec4 albedo = unpackUnorm4x8(cell_data.data[cell_index].albedo);
+
+#ifdef MODE_COMPUTE_LIGHT
+
+	vec3 pos = vec3(posu) + vec3(0.5);
+
+	vec3 emission = vec3(ivec3(cell_data.data[cell_index].emission & 0x3FF, (cell_data.data[cell_index].emission >> 10) & 0x7FF, cell_data.data[cell_index].emission >> 21)) * params.emission_scale;
+	vec4 normal = unpackSnorm4x8(cell_data.data[cell_index].normal);
+
+#ifdef MODE_ANISOTROPIC
+	vec3 accum[6] = vec3[](vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0));
+	const vec3 accum_dirs[6] = vec3[](vec3(1.0, 0.0, 0.0), vec3(-1.0, 0.0, 0.0), vec3(0.0, 1.0, 0.0), vec3(0.0, -1.0, 0.0), vec3(0.0, 0.0, 1.0), vec3(0.0, 0.0, -1.0));
+#else
+	vec3 accum = vec3(0.0);
+#endif
+
+	for (uint i = 0; i < params.light_count; i++) {
+
+		float attenuation;
+		vec3 light_pos;
+
+		if (!compute_light_vector(i, cell_index, pos, attenuation, light_pos)) {
+			continue;
+		}
+
+		vec3 light_dir = pos - light_pos;
+		float distance = length(light_dir);
+		light_dir = normalize(light_dir);
+
+		if (length(normal.xyz) > 0.2 && dot(normal.xyz, light_dir) >= 0) {
+			continue; //not facing the light
+		}
+
+		if (lights.data[i].has_shadow) {
+
+			float distance_adv = get_normal_advance(light_dir);
+
+			distance += distance_adv - mod(distance, distance_adv); //make it reach the center of the box always
+
+			vec3 from = pos - light_dir * distance; //approximate
+			from -= sign(light_dir) * 0.45; //go near the edge towards the light direction to avoid self occlusion
+
+			uint result = raymarch(distance, distance_adv, from, light_dir);
+
+			if (result != cell_index) {
+				continue; //was occluded
+			}
+		}
+
+		vec3 light = lights.data[i].color * albedo.rgb * attenuation * lights.data[i].energy;
+
+#ifdef MODE_ANISOTROPIC
+		for (uint j = 0; j < 6; j++) {
+			accum[j] += max(0.0, dot(accum_dir, -light_dir)) * light + emission;
+		}
+#else
+		if (length(normal.xyz) > 0.2) {
+			accum += max(0.0, dot(normal.xyz, -light_dir)) * light + emission;
+		} else {
+			//all directions
+			accum += light + emission;
+		}
+#endif
+	}
+
+#ifdef MODE_ANISOTROPIC
+
+	output.data[cell_index * 6 + 0] = vec4(accum[0], 0.0);
+	output.data[cell_index * 6 + 1] = vec4(accum[1], 0.0);
+	output.data[cell_index * 6 + 2] = vec4(accum[2], 0.0);
+	output.data[cell_index * 6 + 3] = vec4(accum[3], 0.0);
+	output.data[cell_index * 6 + 4] = vec4(accum[4], 0.0);
+	output.data[cell_index * 6 + 5] = vec4(accum[5], 0.0);
+#else
+	output.data[cell_index] = vec4(accum, 0.0);
+
+#endif
+
+#endif //MODE_COMPUTE_LIGHT
+
+#ifdef MODE_UPDATE_MIPMAPS
+
+	{
+#ifdef MODE_ANISOTROPIC
+		vec3 light_accum[6] = vec3[](vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0));
+#else
+		vec3 light_accum = vec3(0.0);
+#endif
+		float count = 0.0;
+		for (uint i = 0; i < 8; i++) {
+			uint child_index = cell_children.data[cell_index].children[i];
+			if (child_index == NO_CHILDREN) {
+				continue;
+			}
+#ifdef MODE_ANISOTROPIC
+			light_accum[1] += output.data[child_index * 6 + 0].rgb;
+			light_accum[2] += output.data[child_index * 6 + 1].rgb;
+			light_accum[3] += output.data[child_index * 6 + 2].rgb;
+			light_accum[4] += output.data[child_index * 6 + 3].rgb;
+			light_accum[5] += output.data[child_index * 6 + 4].rgb;
+			light_accum[6] += output.data[child_index * 6 + 5].rgb;
+
+#else
+			light_accum += output.data[child_index].rgb;
+
+#endif
+
+			count += 1.0;
+		}
+
+		float divisor = mix(8.0, count, params.propagation);
+#ifdef MODE_ANISOTROPIC
+		output.data[cell_index * 6 + 0] = vec4(light_accum[0] / divisor, 0.0);
+		output.data[cell_index * 6 + 1] = vec4(light_accum[1] / divisor, 0.0);
+		output.data[cell_index * 6 + 2] = vec4(light_accum[2] / divisor, 0.0);
+		output.data[cell_index * 6 + 3] = vec4(light_accum[3] / divisor, 0.0);
+		output.data[cell_index * 6 + 4] = vec4(light_accum[4] / divisor, 0.0);
+		output.data[cell_index * 6 + 5] = vec4(light_accum[5] / divisor, 0.0);
+
+#else
+		output.data[cell_index] = vec4(light_accum / divisor, 0.0);
+#endif
+	}
+#endif
+
+#ifdef MODE_WRITE_TEXTURE
+	{
+	}
+#endif
+}

servers/rendering/rasterizer_rd/shaders/luminance_reduce.glsl

@@ -0,0 +1,87 @@
+/* clang-format off */
+[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+#define BLOCK_SIZE 8
+
+layout(local_size_x = BLOCK_SIZE, local_size_y = BLOCK_SIZE, local_size_z = 1) in;
+/* clang-format on */
+
+shared float tmp_data[BLOCK_SIZE * BLOCK_SIZE];
+
+#ifdef READ_TEXTURE
+
+//use for main texture
+layout(set = 0, binding = 0) uniform sampler2D source_texture;
+
+#else
+
+//use for intermediate textures
+layout(r32f, set = 0, binding = 0) uniform restrict readonly image2D source_luminance;
+
+#endif
+
+layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D dest_luminance;
+
+#ifdef WRITE_LUMINANCE
+layout(set = 2, binding = 0) uniform sampler2D prev_luminance;
+#endif
+
+layout(push_constant, binding = 1, std430) uniform Params {
+	ivec2 source_size;
+	float max_luminance;
+	float min_luminance;
+	float exposure_adjust;
+	float pad[3];
+}
+params;
+
+void main() {
+
+	uint t = gl_LocalInvocationID.y * BLOCK_SIZE + gl_LocalInvocationID.x;
+	ivec2 pos = ivec2(gl_GlobalInvocationID.xy);
+
+	if (any(lessThan(pos, params.source_size))) {
+
+#ifdef READ_TEXTURE
+		vec3 v = texelFetch(source_texture, pos, 0).rgb;
+		tmp_data[t] = max(v.r, max(v.g, v.b));
+#else
+		tmp_data[t] = imageLoad(source_luminance, pos).r;
+#endif
+	} else {
+		tmp_data[t] = 0.0;
+	}
+
+	groupMemoryBarrier();
+	barrier();
+
+	uint size = (BLOCK_SIZE * BLOCK_SIZE) >> 1;
+
+	do {
+		if (t < size) {
+			tmp_data[t] += tmp_data[t + size];
+		}
+		groupMemoryBarrier();
+		barrier();
+
+		size >>= 1;
+
+	} while (size >= 1);
+
+	if (t == 0) {
+		//compute rect size
+		ivec2 rect_size = min(params.source_size - pos, ivec2(BLOCK_SIZE));
+		float avg = tmp_data[0] / float(rect_size.x * rect_size.y);
+		//float avg = tmp_data[0] / float(BLOCK_SIZE*BLOCK_SIZE);
+		pos /= ivec2(BLOCK_SIZE);
+#ifdef WRITE_LUMINANCE
+		float prev_lum = texelFetch(prev_luminance, ivec2(0, 0), 0).r; //1 pixel previous exposure
+		avg = clamp(prev_lum + (avg - prev_lum) * params.exposure_adjust, params.min_luminance, params.max_luminance);
+#endif
+		imageStore(dest_luminance, pos, vec4(avg));
+	}
+}

servers/rendering/rasterizer_rd/shaders/roughness_limiter.glsl

@@ -0,0 +1,73 @@
+/* clang-format off */
+[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
+/* clang-format on */
+
+layout(set = 0, binding = 0) uniform sampler2D source_normal;
+layout(r8, set = 1, binding = 0) uniform restrict writeonly image2D dest_roughness;
+
+layout(push_constant, binding = 1, std430) uniform Params {
+	ivec2 screen_size;
+	float curve;
+	uint pad;
+}
+params;
+
+#define HALF_PI 1.5707963267948966
+
+void main() {
+
+	// Pixel being shaded
+	ivec2 pos = ivec2(gl_GlobalInvocationID.xy);
+	if (any(greaterThan(pos, params.screen_size))) { //too large, do nothing
+		return;
+	}
+
+	vec3 normal_accum = vec3(0.0);
+	float accum = 0.0;
+	for (int i = 0; i <= 1; i++) {
+		for (int j = 0; j <= 1; j++) {
+			normal_accum += normalize(texelFetch(source_normal, pos + ivec2(i, j), 0).xyz * 2.0 - 1.0);
+			accum += 1.0;
+		}
+	}
+
+	normal_accum /= accum;
+
+	float r = length(normal_accum);
+
+	float limit;
+
+	if (r < 1.0) {
+		float threshold = 0.4;
+
+		/*
+		//Formula from Filament, does not make sense to me.
+
+		float r2 = r * r;
+		float kappa = (3.0f * r - r * r2) / (1.0f - r2);
+		float variance = 0.25f / kappa;
+		limit = sqrt(min(2.0f * variance, threshold * threshold));
+//*/
+		/*
+		//Formula based on probability distribution graph
+
+		float width = acos(max(0.0,r)); // convert to angle (width)
+		float roughness = pow(width,1.7)*0.854492; //approximate (crappy) formula to convert to roughness
+		limit = min(sqrt(roughness), threshold); //convert to perceptual roughness and apply threshold
+//*/
+
+		limit = min(sqrt(pow(acos(max(0.0, r)) / HALF_PI, params.curve)), threshold); //convert to perceptual roughness and apply threshold
+
+		//limit = 0.5;
+	} else {
+		limit = 0.0;
+	}
+
+	imageStore(dest_roughness, pos, vec4(limit));
+}

servers/rendering/rasterizer_rd/shaders/scene_high_end_inc.glsl

@@ -0,0 +1,266 @@
+#define M_PI 3.14159265359
+#define ROUGHNESS_MAX_LOD 5
+
+layout(push_constant, binding = 0, std430) uniform DrawCall {
+	uint instance_index;
+	uint pad[3]; //16 bits minimum size
+}
+draw_call;
+
+/* Set 0 Scene data that never changes, ever */
+
+#define SAMPLER_NEAREST_CLAMP 0
+#define SAMPLER_LINEAR_CLAMP 1
+#define SAMPLER_NEAREST_WITH_MIPMAPS_CLAMP 2
+#define SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP 3
+#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_CLAMP 4
+#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_CLAMP 5
+#define SAMPLER_NEAREST_REPEAT 6
+#define SAMPLER_LINEAR_REPEAT 7
+#define SAMPLER_NEAREST_WITH_MIPMAPS_REPEAT 8
+#define SAMPLER_LINEAR_WITH_MIPMAPS_REPEAT 9
+#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_REPEAT 10
+#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_REPEAT 11
+
+layout(set = 0, binding = 1) uniform sampler material_samplers[12];
+
+layout(set = 0, binding = 2) uniform sampler shadow_sampler;
+
+layout(set = 0, binding = 3, std140) uniform SceneData {
+
+	mat4 projection_matrix;
+	mat4 inv_projection_matrix;
+
+	mat4 camera_matrix;
+	mat4 inv_camera_matrix;
+
+	vec2 viewport_size;
+	vec2 screen_pixel_size;
+
+	//used for shadow mapping only
+	float z_offset;
+	float z_slope_scale;
+
+	float time;
+	float reflection_multiplier; // one normally, zero when rendering reflections
+
+	vec4 ambient_light_color_energy;
+
+	float ambient_color_sky_mix;
+	bool use_ambient_light;
+	bool use_ambient_cubemap;
+	bool use_reflection_cubemap;
+
+	mat3 radiance_inverse_xform;
+
+	vec2 shadow_atlas_pixel_size;
+	vec2 directional_shadow_pixel_size;
+
+	uint directional_light_count;
+	float dual_paraboloid_side;
+	float z_far;
+	float z_near;
+
+	bool ssao_enabled;
+	float ssao_light_affect;
+	float ssao_ao_affect;
+	bool roughness_limiter_enabled;
+
+	vec4 ao_color;
+
+#if 0
+	vec4 ambient_light_color;
+	vec4 bg_color;
+
+	vec4 fog_color_enabled;
+	vec4 fog_sun_color_amount;
+
+	float ambient_energy;
+	float bg_energy;
+#endif
+
+#if 0
+	vec2 shadow_atlas_pixel_size;
+	vec2 directional_shadow_pixel_size;
+
+	float z_far;
+
+	float subsurface_scatter_width;
+	float ambient_occlusion_affect_light;
+	float ambient_occlusion_affect_ao_channel;
+	float opaque_prepass_threshold;
+
+	bool fog_depth_enabled;
+	float fog_depth_begin;
+	float fog_depth_end;
+	float fog_density;
+	float fog_depth_curve;
+	bool fog_transmit_enabled;
+	float fog_transmit_curve;
+	bool fog_height_enabled;
+	float fog_height_min;
+	float fog_height_max;
+	float fog_height_curve;
+#endif
+}
+scene_data;
+
+#define INSTANCE_FLAGS_FORWARD_MASK 0x7
+#define INSTANCE_FLAGS_FORWARD_OMNI_LIGHT_SHIFT 3
+#define INSTANCE_FLAGS_FORWARD_SPOT_LIGHT_SHIFT 6
+#define INSTANCE_FLAGS_FORWARD_DECAL_SHIFT 9
+
+#define INSTANCE_FLAGS_MULTIMESH (1 << 12)
+#define INSTANCE_FLAGS_MULTIMESH_FORMAT_2D (1 << 13)
+#define INSTANCE_FLAGS_MULTIMESH_HAS_COLOR (1 << 14)
+#define INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA (1 << 15)
+#define INSTANCE_FLAGS_MULTIMESH_STRIDE_SHIFT 16
+//3 bits of stride
+#define INSTANCE_FLAGS_MULTIMESH_STRIDE_MASK 0x7
+
+#define INSTANCE_FLAGS_SKELETON (1 << 19)
+
+struct InstanceData {
+	mat4 transform;
+	mat4 normal_transform;
+	uint flags;
+	uint instance_ofs; //instance_offset in instancing/skeleton buffer
+	uint gi_offset; //GI information when using lightmapping (VCT or lightmap)
+	uint layer_mask;
+};
+
+layout(set = 0, binding = 4, std430) buffer Instances {
+	InstanceData data[];
+}
+instances;
+
+struct LightData { //this structure needs to be 128 bits
+	vec3 position;
+	float inv_radius;
+	vec3 direction;
+	uint attenuation_energy; //attenuation
+	uint color_specular; //rgb color, a specular (8 bit unorm)
+	uint cone_attenuation_angle; // attenuation and angle, (16bit float)
+	uint mask;
+	uint shadow_color_enabled; //shadow rgb color, a>0.5 enabled (8bit unorm)
+	vec4 atlas_rect; //used for shadow atlas uv on omni, and for projection atlas on spot
+	mat4 shadow_matrix;
+};
+
+layout(set = 0, binding = 5, std140) uniform Lights {
+	LightData data[MAX_LIGHT_DATA_STRUCTS];
+}
+lights;
+
+struct ReflectionData {
+
+	vec3 box_extents;
+	float index;
+	vec3 box_offset;
+	uint mask;
+	vec4 params; // intensity, 0, interior , boxproject
+	vec4 ambient; // ambient color, energy
+	mat4 local_matrix; // up to here for spot and omni, rest is for directional
+	// notes: for ambientblend, use distance to edge to blend between already existing global environment
+};
+
+layout(set = 0, binding = 6, std140) uniform ReflectionProbeData {
+	ReflectionData data[MAX_REFLECTION_DATA_STRUCTS];
+}
+reflections;
+
+struct DirectionalLightData {
+	vec3 direction;
+	float energy;
+	vec3 color;
+	float specular;
+	vec3 shadow_color;
+	uint mask;
+	bool blend_splits;
+	bool shadow_enabled;
+	float fade_from;
+	float fade_to;
+	vec4 shadow_split_offsets;
+	mat4 shadow_matrix1;
+	mat4 shadow_matrix2;
+	mat4 shadow_matrix3;
+	mat4 shadow_matrix4;
+};
+
+layout(set = 0, binding = 7, std140) uniform DirectionalLights {
+	DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS];
+}
+directional_lights;
+
+struct GIProbeData {
+	mat4 xform;
+	vec3 bounds;
+	float dynamic_range;
+
+	float bias;
+	float normal_bias;
+	bool blend_ambient;
+	uint texture_slot;
+
+	float anisotropy_strength;
+	float ambient_occlusion;
+	float ambient_occlusion_size;
+	uint pad2;
+};
+
+layout(set = 0, binding = 8, std140) uniform GIProbes {
+	GIProbeData data[MAX_GI_PROBES];
+}
+gi_probes;
+
+layout(set = 0, binding = 9) uniform texture3D gi_probe_textures[MAX_GI_PROBE_TEXTURES];
+
+#define CLUSTER_COUNTER_SHIFT 20
+#define CLUSTER_POINTER_MASK ((1 << CLUSTER_COUNTER_SHIFT) - 1)
+#define CLUSTER_COUNTER_MASK 0xfff
+
+layout(set = 0, binding = 10) uniform utexture3D cluster_texture;
+
+layout(set = 0, binding = 11, std430) buffer ClusterData {
+	uint indices[];
+}
+cluster_data;
+
+layout(set = 0, binding = 12) uniform texture2D directional_shadow_atlas;
+
+// decal atlas
+
+/* Set 1, Radiance */
+
+#ifdef USE_RADIANCE_CUBEMAP_ARRAY
+
+layout(set = 1, binding = 0) uniform textureCubeArray radiance_cubemap;
+
+#else
+
+layout(set = 1, binding = 0) uniform textureCube radiance_cubemap;
+
+#endif
+
+/* Set 2, Reflection and Shadow Atlases (view dependant) */
+
+layout(set = 2, binding = 0) uniform textureCubeArray reflection_atlas;
+
+layout(set = 2, binding = 1) uniform texture2D shadow_atlas;
+
+/* Set 1, Render Buffers */
+
+layout(set = 3, binding = 0) uniform texture2D depth_buffer;
+layout(set = 3, binding = 1) uniform texture2D color_buffer;
+layout(set = 3, binding = 2) uniform texture2D normal_buffer;
+layout(set = 3, binding = 3) uniform texture2D roughness_buffer;
+layout(set = 3, binding = 4) uniform texture2D ao_buffer;
+
+/* Set 4 Skeleton & Instancing (Multimesh) */
+
+layout(set = 4, binding = 0, std430) buffer Transforms {
+	vec4 data[];
+}
+transforms;
+
+/* Set 5 User Material */

servers/rendering/rasterizer_rd/shaders/sky.glsl

@@ -0,0 +1,181 @@
+/* clang-format off */
+[vertex]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(location = 0) out vec2 uv_interp;
+/* clang-format on */
+
+layout(push_constant, binding = 1, std430) uniform Params {
+	mat3 orientation;
+	vec4 proj;
+	vec4 position_multiplier;
+	float time;
+}
+params;
+
+void main() {
+
+	vec2 base_arr[4] = vec2[](vec2(-1.0, -1.0), vec2(-1.0, 1.0), vec2(1.0, 1.0), vec2(1.0, -1.0));
+	uv_interp = base_arr[gl_VertexIndex];
+	gl_Position = vec4(uv_interp, 1.0, 1.0);
+}
+
+/* clang-format off */
+[fragment]
+
+#version 450
+
+VERSION_DEFINES
+
+#define M_PI 3.14159265359
+
+layout(location = 0) in vec2 uv_interp;
+/* clang-format on */
+
+layout(push_constant, binding = 1, std430) uniform Params {
+	mat3 orientation;
+	vec4 proj;
+	vec4 position_multiplier;
+	float time; //TODO consider adding vec2 screen res, and float radiance size
+}
+params;
+
+#define SAMPLER_NEAREST_CLAMP 0
+#define SAMPLER_LINEAR_CLAMP 1
+#define SAMPLER_NEAREST_WITH_MIPMAPS_CLAMP 2
+#define SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP 3
+#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_CLAMP 4
+#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_CLAMP 5
+#define SAMPLER_NEAREST_REPEAT 6
+#define SAMPLER_LINEAR_REPEAT 7
+#define SAMPLER_NEAREST_WITH_MIPMAPS_REPEAT 8
+#define SAMPLER_LINEAR_WITH_MIPMAPS_REPEAT 9
+#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_REPEAT 10
+#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_REPEAT 11
+
+layout(set = 0, binding = 0) uniform sampler material_samplers[12];
+
+#ifdef USE_MATERIAL_UNIFORMS
+layout(set = 1, binding = 0, std140) uniform MaterialUniforms{
+	/* clang-format off */
+
+MATERIAL_UNIFORMS
+
+	/* clang-format on */
+} material;
+#endif
+
+layout(set = 2, binding = 0) uniform textureCube radiance;
+#ifdef USE_CUBEMAP_PASS
+layout(set = 2, binding = 1) uniform textureCube half_res;
+layout(set = 2, binding = 2) uniform textureCube quarter_res;
+#else
+layout(set = 2, binding = 1) uniform texture2D half_res;
+layout(set = 2, binding = 2) uniform texture2D quarter_res;
+#endif
+
+#ifdef USE_CUBEMAP_PASS
+#define AT_CUBEMAP_PASS true
+#else
+#define AT_CUBEMAP_PASS false
+#endif
+
+#ifdef USE_HALF_RES_PASS
+#define AT_HALF_RES_PASS true
+#else
+#define AT_HALF_RES_PASS false
+#endif
+
+#ifdef USE_QUARTER_RES_PASS
+#define AT_QUARTER_RES_PASS true
+#else
+#define AT_QUARTER_RES_PASS false
+#endif
+
+struct DirectionalLightData {
+	vec3 direction;
+	float energy;
+	vec3 color;
+	bool enabled;
+};
+
+layout(set = 3, binding = 0, std140) uniform DirectionalLights {
+	DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS];
+}
+directional_lights;
+
+/* clang-format off */
+
+FRAGMENT_SHADER_GLOBALS
+
+/* clang-format on */
+
+layout(location = 0) out vec4 frag_color;
+
+void main() {
+
+	vec3 cube_normal;
+	cube_normal.z = -1.0;
+	cube_normal.x = (cube_normal.z * (-uv_interp.x - params.proj.x)) / params.proj.y;
+	cube_normal.y = -(cube_normal.z * (-uv_interp.y - params.proj.z)) / params.proj.w;
+	cube_normal = mat3(params.orientation) * cube_normal;
+	cube_normal.z = -cube_normal.z;
+	cube_normal = normalize(cube_normal);
+
+	vec2 uv = uv_interp * 0.5 + 0.5;
+
+	vec2 panorama_coords = vec2(atan(cube_normal.x, cube_normal.z), acos(cube_normal.y));
+
+	if (panorama_coords.x < 0.0) {
+		panorama_coords.x += M_PI * 2.0;
+	}
+
+	panorama_coords /= vec2(M_PI * 2.0, M_PI);
+
+	vec3 color = vec3(0.0, 0.0, 0.0);
+	float alpha = 1.0; // Only available to subpasses
+	vec4 half_res_color = vec4(1.0);
+	vec4 quarter_res_color = vec4(1.0);
+
+#ifdef USE_CUBEMAP_PASS
+	float using_cubemap = 1.0;
+#ifdef USES_HALF_RES_COLOR
+	half_res_color = texture(samplerCube(half_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), cube_normal);
+#endif
+#ifdef USES_QUARTER_RES_COLOR
+	quarter_res_color = texture(samplerCube(quarter_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), cube_normal);
+#endif
+#else
+	float using_cubemap = 0.0;
+#ifdef USES_HALF_RES_COLOR
+	half_res_color = textureLod(sampler2D(half_res, material_samplers[SAMPLER_LINEAR_CLAMP]), uv, 0.0);
+#endif
+#ifdef USES_QUARTER_RES_COLOR
+	quarter_res_color = textureLod(sampler2D(quarter_res, material_samplers[SAMPLER_LINEAR_CLAMP]), uv, 0.0);
+#endif
+#endif
+
+// unused, just here to make our compiler happy, make sure we don't execute any light code the user adds in..
+#ifndef REALLYINCLUDETHIS
+	{
+		/* clang-format off */
+
+LIGHT_SHADER_CODE
+
+		/* clang-format on */
+	}
+#endif
+	{
+		/* clang-format off */
+
+FRAGMENT_SHADER_CODE
+
+		/* clang-format on */
+	}
+
+	frag_color.rgb = color * params.position_multiplier.w;
+	frag_color.a = alpha;
+}

servers/rendering/rasterizer_rd/shaders/ssao.glsl

@@ -0,0 +1,252 @@
+/* clang-format off */
+[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
+/* clang-format on */
+
+#define TWO_PI 6.283185307179586476925286766559
+
+#ifdef SSAO_QUALITY_HIGH
+#define NUM_SAMPLES (20)
+#endif
+
+#ifdef SSAO_QUALITY_ULTRA
+#define NUM_SAMPLES (48)
+#endif
+
+#ifdef SSAO_QUALITY_LOW
+#define NUM_SAMPLES (8)
+#endif
+
+#if !defined(SSAO_QUALITY_LOW) && !defined(SSAO_QUALITY_HIGH) && !defined(SSAO_QUALITY_ULTRA)
+#define NUM_SAMPLES (12)
+#endif
+
+// If using depth mip levels, the log of the maximum pixel offset before we need to switch to a lower
+// miplevel to maintain reasonable spatial locality in the cache
+// If this number is too small (< 3), too many taps will land in the same pixel, and we'll get bad variance that manifests as flashing.
+// If it is too high (> 5), we'll get bad performance because we're not using the MIP levels effectively
+#define LOG_MAX_OFFSET (3)
+
+// This must be less than or equal to the MAX_MIP_LEVEL defined in SSAO.cpp
+#define MAX_MIP_LEVEL (4)
+
+// This is the number of turns around the circle that the spiral pattern makes.  This should be prime to prevent
+// taps from lining up.  This particular choice was tuned for NUM_SAMPLES == 9
+
+const int ROTATIONS[] = int[](
+		1, 1, 2, 3, 2, 5, 2, 3, 2,
+		3, 3, 5, 5, 3, 4, 7, 5, 5, 7,
+		9, 8, 5, 5, 7, 7, 7, 8, 5, 8,
+		11, 12, 7, 10, 13, 8, 11, 8, 7, 14,
+		11, 11, 13, 12, 13, 19, 17, 13, 11, 18,
+		19, 11, 11, 14, 17, 21, 15, 16, 17, 18,
+		13, 17, 11, 17, 19, 18, 25, 18, 19, 19,
+		29, 21, 19, 27, 31, 29, 21, 18, 17, 29,
+		31, 31, 23, 18, 25, 26, 25, 23, 19, 34,
+		19, 27, 21, 25, 39, 29, 17, 21, 27);
+/* clang-format on */
+
+//#define NUM_SPIRAL_TURNS (7)
+const int NUM_SPIRAL_TURNS = ROTATIONS[NUM_SAMPLES - 1];
+
+layout(set = 0, binding = 0) uniform sampler2D source_depth_mipmaps;
+layout(r8, set = 1, binding = 0) uniform restrict writeonly image2D dest_image;
+
+#ifndef USE_HALF_SIZE
+layout(set = 2, binding = 0) uniform sampler2D source_depth;
+#endif
+
+layout(set = 3, binding = 0) uniform sampler2D source_normal;
+
+layout(push_constant, binding = 1, std430) uniform Params {
+	ivec2 screen_size;
+	float z_far;
+	float z_near;
+
+	bool orthogonal;
+	float intensity_div_r6;
+	float radius;
+	float bias;
+
+	vec4 proj_info;
+	vec2 pixel_size;
+	float proj_scale;
+	uint pad;
+}
+params;
+
+vec3 reconstructCSPosition(vec2 S, float z) {
+	if (params.orthogonal) {
+		return vec3((S.xy * params.proj_info.xy + params.proj_info.zw), z);
+	} else {
+		return vec3((S.xy * params.proj_info.xy + params.proj_info.zw) * z, z);
+	}
+}
+
+vec3 getPosition(ivec2 ssP) {
+	vec3 P;
+#ifdef USE_HALF_SIZE
+	P.z = texelFetch(source_depth_mipmaps, ssP, 0).r;
+	P.z = -P.z;
+#else
+	P.z = texelFetch(source_depth, ssP, 0).r;
+
+	P.z = P.z * 2.0 - 1.0;
+	if (params.orthogonal) {
+		P.z = ((P.z + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0;
+	} else {
+		P.z = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - P.z * (params.z_far - params.z_near));
+	}
+	P.z = -P.z;
+#endif
+	// Offset to pixel center
+	P = reconstructCSPosition(vec2(ssP) + vec2(0.5), P.z);
+	return P;
+}
+
+/** Returns a unit vector and a screen-space radius for the tap on a unit disk (the caller should scale by the actual disk radius) */
+vec2 tapLocation(int sampleNumber, float spinAngle, out float ssR) {
+	// Radius relative to ssR
+	float alpha = (float(sampleNumber) + 0.5) * (1.0 / float(NUM_SAMPLES));
+	float angle = alpha * (float(NUM_SPIRAL_TURNS) * 6.28) + spinAngle;
+
+	ssR = alpha;
+	return vec2(cos(angle), sin(angle));
+}
+
+/** Read the camera-space position of the point at screen-space pixel ssP + unitOffset * ssR.  Assumes length(unitOffset) == 1 */
+vec3 getOffsetPosition(ivec2 ssP, float ssR) {
+	// Derivation:
+	//  mipLevel = floor(log(ssR / MAX_OFFSET));
+
+	int mipLevel = clamp(int(floor(log2(ssR))) - LOG_MAX_OFFSET, 0, MAX_MIP_LEVEL);
+
+	vec3 P;
+
+	// We need to divide by 2^mipLevel to read the appropriately scaled coordinate from a MIP-map.
+	// Manually clamp to the texture size because texelFetch bypasses the texture unit
+	ivec2 mipP = clamp(ssP >> mipLevel, ivec2(0), (params.screen_size >> mipLevel) - ivec2(1));
+
+#ifdef USE_HALF_SIZE
+	P.z = texelFetch(source_depth_mipmaps, mipP, mipLevel).r;
+	P.z = -P.z;
+#else
+	if (mipLevel < 1) {
+		//read from depth buffer
+		P.z = texelFetch(source_depth, mipP, 0).r;
+		P.z = P.z * 2.0 - 1.0;
+		if (params.orthogonal) {
+			P.z = ((P.z + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0;
+		} else {
+			P.z = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - P.z * (params.z_far - params.z_near));
+		}
+		P.z = -P.z;
+
+	} else {
+		//read from mipmaps
+		P.z = texelFetch(source_depth_mipmaps, mipP, mipLevel - 1).r;
+		P.z = -P.z;
+	}
+#endif
+
+	// Offset to pixel center
+	P = reconstructCSPosition(vec2(ssP) + vec2(0.5), P.z);
+
+	return P;
+}
+
+/** Compute the occlusion due to sample with index \a i about the pixel at \a ssC that corresponds
+	to camera-space point \a C with unit normal \a n_C, using maximum screen-space sampling radius \a ssDiskRadius
+
+	Note that units of H() in the HPG12 paper are meters, not
+	unitless.  The whole falloff/sampling function is therefore
+	unitless.  In this implementation, we factor out (9 / radius).
+
+	Four versions of the falloff function are implemented below
+*/
+float sampleAO(in ivec2 ssC, in vec3 C, in vec3 n_C, in float ssDiskRadius, in float p_radius, in int tapIndex, in float randomPatternRotationAngle) {
+	// Offset on the unit disk, spun for this pixel
+	float ssR;
+	vec2 unitOffset = tapLocation(tapIndex, randomPatternRotationAngle, ssR);
+	ssR *= ssDiskRadius;
+
+	ivec2 ssP = ivec2(ssR * unitOffset) + ssC;
+
+	if (any(lessThan(ssP, ivec2(0))) || any(greaterThanEqual(ssP, params.screen_size))) {
+		return 0.0;
+	}
+
+	// The occluding point in camera space
+	vec3 Q = getOffsetPosition(ssP, ssR);
+
+	vec3 v = Q - C;
+
+	float vv = dot(v, v);
+	float vn = dot(v, n_C);
+
+	const float epsilon = 0.01;
+	float radius2 = p_radius * p_radius;
+
+	// A: From the HPG12 paper
+	// Note large epsilon to avoid overdarkening within cracks
+	//return float(vv < radius2) * max((vn - bias) / (epsilon + vv), 0.0) * radius2 * 0.6;
+
+	// B: Smoother transition to zero (lowers contrast, smoothing out corners). [Recommended]
+	float f = max(radius2 - vv, 0.0);
+	return f * f * f * max((vn - params.bias) / (epsilon + vv), 0.0);
+
+	// C: Medium contrast (which looks better at high radii), no division.  Note that the
+	// contribution still falls off with radius^2, but we've adjusted the rate in a way that is
+	// more computationally efficient and happens to be aesthetically pleasing.
+	// return 4.0 * max(1.0 - vv * invRadius2, 0.0) * max(vn - bias, 0.0);
+
+	// D: Low contrast, no division operation
+	// return 2.0 * float(vv < radius * radius) * max(vn - bias, 0.0);
+}
+
+void main() {
+	// Pixel being shaded
+	ivec2 ssC = ivec2(gl_GlobalInvocationID.xy);
+	if (any(greaterThan(ssC, params.screen_size))) { //too large, do nothing
+		return;
+	}
+
+	// World space point being shaded
+	vec3 C = getPosition(ssC);
+
+#ifdef USE_HALF_SIZE
+	vec3 n_C = texelFetch(source_normal, ssC << 1, 0).xyz * 2.0 - 1.0;
+#else
+	vec3 n_C = texelFetch(source_normal, ssC, 0).xyz * 2.0 - 1.0;
+#endif
+	n_C = normalize(n_C);
+	n_C.y = -n_C.y; //because this code reads flipped
+
+	// Hash function used in the HPG12 AlchemyAO paper
+	float randomPatternRotationAngle = mod(float((3 * ssC.x ^ ssC.y + ssC.x * ssC.y) * 10), TWO_PI);
+
+	// Reconstruct normals from positions. These will lead to 1-pixel black lines
+	// at depth discontinuities, however the blur will wipe those out so they are not visible
+	// in the final image.
+
+	// Choose the screen-space sample radius
+	// proportional to the projected area of the sphere
+
+	float ssDiskRadius = -params.proj_scale * params.radius;
+	if (!params.orthogonal) {
+		ssDiskRadius = -params.proj_scale * params.radius / C.z;
+	}
+	float sum = 0.0;
+	for (int i = 0; i < NUM_SAMPLES; ++i) {
+		sum += sampleAO(ssC, C, n_C, ssDiskRadius, params.radius, i, randomPatternRotationAngle);
+	}
+
+	float A = max(0.0, 1.0 - sum * params.intensity_div_r6 * (5.0 / float(NUM_SAMPLES)));
+
+	imageStore(dest_image, ssC, vec4(A));
+}

servers/rendering/rasterizer_rd/shaders/ssao_blur.glsl

@@ -0,0 +1,157 @@
+/* clang-format off */
+[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
+/* clang-format on */
+
+layout(set = 0, binding = 0) uniform sampler2D source_ssao;
+layout(set = 1, binding = 0) uniform sampler2D source_depth;
+#ifdef MODE_UPSCALE
+layout(set = 2, binding = 0) uniform sampler2D source_depth_mipmaps;
+#endif
+
+layout(r8, set = 3, binding = 0) uniform restrict writeonly image2D dest_image;
+
+//////////////////////////////////////////////////////////////////////////////////////////////
+// Tunable Parameters:
+
+layout(push_constant, binding = 1, std430) uniform Params {
+	float edge_sharpness; /** Increase to make depth edges crisper. Decrease to reduce flicker. */
+	int filter_scale;
+	float z_far;
+	float z_near;
+	bool orthogonal;
+	uint pad0;
+	uint pad1;
+	uint pad2;
+	ivec2 axis; /** (1, 0) or (0, 1) */
+	ivec2 screen_size;
+}
+params;
+
+/** Filter radius in pixels. This will be multiplied by SCALE. */
+#define R (4)
+
+//////////////////////////////////////////////////////////////////////////////////////////////
+
+// Gaussian coefficients
+const float gaussian[R + 1] =
+		//float[](0.356642, 0.239400, 0.072410, 0.009869);
+		//float[](0.398943, 0.241971, 0.053991, 0.004432, 0.000134);  // stddev = 1.0
+		float[](0.153170, 0.144893, 0.122649, 0.092902, 0.062970); // stddev = 2.0
+//float[](0.111220, 0.107798, 0.098151, 0.083953, 0.067458, 0.050920, 0.036108); // stddev = 3.0
+
+void main() {
+
+	// Pixel being shaded
+	ivec2 ssC = ivec2(gl_GlobalInvocationID.xy);
+	if (any(greaterThan(ssC, params.screen_size))) { //too large, do nothing
+		return;
+	}
+
+#ifdef MODE_UPSCALE
+
+	//closest one should be the same pixel, but check nearby just in case
+	float depth = texelFetch(source_depth, ssC, 0).r;
+
+	depth = depth * 2.0 - 1.0;
+	if (params.orthogonal) {
+		depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0;
+	} else {
+		depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near));
+	}
+
+	vec2 pixel_size = 1.0 / vec2(params.screen_size);
+	vec2 closest_uv = vec2(ssC) * pixel_size + pixel_size * 0.5;
+	vec2 from_uv = closest_uv;
+	vec2 ps2 = pixel_size; // * 2.0;
+
+	float closest_depth = abs(textureLod(source_depth_mipmaps, closest_uv, 0.0).r - depth);
+
+	vec2 offsets[4] = vec2[](vec2(ps2.x, 0), vec2(-ps2.x, 0), vec2(0, ps2.y), vec2(0, -ps2.y));
+	for (int i = 0; i < 4; i++) {
+		vec2 neighbour = from_uv + offsets[i];
+		float neighbour_depth = abs(textureLod(source_depth_mipmaps, neighbour, 0.0).r - depth);
+		if (neighbour_depth < closest_depth) {
+			closest_uv = neighbour;
+			closest_depth = neighbour_depth;
+		}
+	}
+
+	float visibility = textureLod(source_ssao, closest_uv, 0.0).r;
+	imageStore(dest_image, ssC, vec4(visibility));
+#else
+
+	float depth = texelFetch(source_depth, ssC, 0).r;
+
+#ifdef MODE_FULL_SIZE
+	depth = depth * 2.0 - 1.0;
+
+	if (params.orthogonal) {
+		depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0;
+	} else {
+		depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near));
+	}
+
+#endif
+	float depth_divide = 1.0 / params.z_far;
+
+	//depth *= depth_divide;
+
+	/*
+	if (depth > params.z_far * 0.999) {
+		discard; //skybox
+	}
+	*/
+
+	float sum = texelFetch(source_ssao, ssC, 0).r;
+
+	// Base weight for depth falloff.  Increase this for more blurriness,
+	// decrease it for better edge discrimination
+	float BASE = gaussian[0];
+	float totalWeight = BASE;
+	sum *= totalWeight;
+
+	ivec2 clamp_limit = params.screen_size - ivec2(1);
+
+	for (int r = -R; r <= R; ++r) {
+		// We already handled the zero case above.  This loop should be unrolled and the static branch optimized out,
+		// so the IF statement has no runtime cost
+		if (r != 0) {
+
+			ivec2 ppos = ssC + params.axis * (r * params.filter_scale);
+			float value = texelFetch(source_ssao, clamp(ppos, ivec2(0), clamp_limit), 0).r;
+			ivec2 rpos = clamp(ppos, ivec2(0), clamp_limit);
+
+			float temp_depth = texelFetch(source_depth, rpos, 0).r;
+#ifdef MODE_FULL_SIZE
+			temp_depth = temp_depth * 2.0 - 1.0;
+			if (params.orthogonal) {
+				temp_depth = ((temp_depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0;
+			} else {
+				temp_depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - temp_depth * (params.z_far - params.z_near));
+			}
+			//temp_depth *= depth_divide;
+#endif
+			// spatial domain: offset gaussian tap
+			float weight = 0.3 + gaussian[abs(r)];
+			//weight *= max(0.0, dot(temp_normal, normal));
+
+			// range domain (the "bilateral" weight). As depth difference increases, decrease weight.
+			weight *= max(0.0, 1.0 - params.edge_sharpness * abs(temp_depth - depth));
+
+			sum += value * weight;
+			totalWeight += weight;
+		}
+	}
+
+	const float epsilon = 0.0001;
+	float visibility = sum / (totalWeight + epsilon);
+
+	imageStore(dest_image, ssC, vec4(visibility));
+#endif
+}

servers/rendering/rasterizer_rd/shaders/ssao_minify.glsl

@@ -0,0 +1,48 @@
+/* clang-format off */
+[compute]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
+/* clang-format on */
+
+layout(push_constant, binding = 1, std430) uniform Params {
+	vec2 pixel_size;
+	float z_far;
+	float z_near;
+	ivec2 source_size;
+	bool orthogonal;
+	uint pad;
+}
+params;
+
+#ifdef MINIFY_START
+layout(set = 0, binding = 0) uniform sampler2D source_texture;
+#else
+layout(r32f, set = 0, binding = 0) uniform restrict readonly image2D source_image;
+#endif
+layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D dest_image;
+
+void main() {
+
+	ivec2 pos = ivec2(gl_GlobalInvocationID.xy);
+
+	if (any(greaterThan(pos, params.source_size >> 1))) { //too large, do nothing
+		return;
+	}
+
+#ifdef MINIFY_START
+	float depth = texelFetch(source_texture, pos << 1, 0).r * 2.0 - 1.0;
+	if (params.orthogonal) {
+		depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0;
+	} else {
+		depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near));
+	}
+#else
+	float depth = imageLoad(source_image, pos << 1).r;
+#endif
+
+	imageStore(dest_image, pos, vec4(depth));
+}

servers/rendering/rasterizer_rd/shaders/tonemap.glsl

@@ -0,0 +1,305 @@
+/* clang-format off */
+[vertex]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(location = 0) out vec2 uv_interp;
+/* clang-format on */
+
+void main() {
+
+	vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0));
+	uv_interp = base_arr[gl_VertexIndex];
+	gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0);
+}
+
+/* clang-format off */
+[fragment]
+
+#version 450
+
+VERSION_DEFINES
+
+layout(location = 0) in vec2 uv_interp;
+/* clang-format on */
+
+layout(set = 0, binding = 0) uniform sampler2D source_color;
+layout(set = 1, binding = 0) uniform sampler2D source_auto_exposure;
+layout(set = 2, binding = 0) uniform sampler2D source_glow;
+layout(set = 3, binding = 0) uniform sampler3D color_correction;
+
+layout(push_constant, binding = 1, std430) uniform Params {
+	vec3 bcs;
+	bool use_bcs;
+
+	bool use_glow;
+	bool use_auto_exposure;
+	bool use_color_correction;
+	uint tonemapper;
+
+	uvec2 glow_texture_size;
+
+	float glow_intensity;
+	uint glow_level_flags;
+	uint glow_mode;
+
+	float exposure;
+	float white;
+	float auto_exposure_grey;
+}
+params;
+
+layout(location = 0) out vec4 frag_color;
+
+#ifdef USE_GLOW_FILTER_BICUBIC
+// w0, w1, w2, and w3 are the four cubic B-spline basis functions
+float w0(float a) {
+	return (1.0f / 6.0f) * (a * (a * (-a + 3.0f) - 3.0f) + 1.0f);
+}
+
+float w1(float a) {
+	return (1.0f / 6.0f) * (a * a * (3.0f * a - 6.0f) + 4.0f);
+}
+
+float w2(float a) {
+	return (1.0f / 6.0f) * (a * (a * (-3.0f * a + 3.0f) + 3.0f) + 1.0f);
+}
+
+float w3(float a) {
+	return (1.0f / 6.0f) * (a * a * a);
+}
+
+// g0 and g1 are the two amplitude functions
+float g0(float a) {
+	return w0(a) + w1(a);
+}
+
+float g1(float a) {
+	return w2(a) + w3(a);
+}
+
+// h0 and h1 are the two offset functions
+float h0(float a) {
+	return -1.0f + w1(a) / (w0(a) + w1(a));
+}
+
+float h1(float a) {
+	return 1.0f + w3(a) / (w2(a) + w3(a));
+}
+
+vec4 texture2D_bicubic(sampler2D tex, vec2 uv, int p_lod) {
+	float lod = float(p_lod);
+	vec2 tex_size = vec2(params.glow_texture_size >> p_lod);
+	vec2 pixel_size = vec2(1.0f) / tex_size;
+
+	uv = uv * tex_size + vec2(0.5f);
+
+	vec2 iuv = floor(uv);
+	vec2 fuv = fract(uv);
+
+	float g0x = g0(fuv.x);
+	float g1x = g1(fuv.x);
+	float h0x = h0(fuv.x);
+	float h1x = h1(fuv.x);
+	float h0y = h0(fuv.y);
+	float h1y = h1(fuv.y);
+
+	vec2 p0 = (vec2(iuv.x + h0x, iuv.y + h0y) - vec2(0.5f)) * pixel_size;
+	vec2 p1 = (vec2(iuv.x + h1x, iuv.y + h0y) - vec2(0.5f)) * pixel_size;
+	vec2 p2 = (vec2(iuv.x + h0x, iuv.y + h1y) - vec2(0.5f)) * pixel_size;
+	vec2 p3 = (vec2(iuv.x + h1x, iuv.y + h1y) - vec2(0.5f)) * pixel_size;
+
+	return (g0(fuv.y) * (g0x * textureLod(tex, p0, lod) + g1x * textureLod(tex, p1, lod))) +
+		   (g1(fuv.y) * (g0x * textureLod(tex, p2, lod) + g1x * textureLod(tex, p3, lod)));
+}
+
+#define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) texture2D_bicubic(m_tex, m_uv, m_lod)
+
+#else
+
+#define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) textureLod(m_tex, m_uv, float(m_lod))
+
+#endif
+
+vec3 tonemap_filmic(vec3 color, float white) {
+	// exposure bias: input scale (color *= bias, white *= bias) to make the brightness consistent with other tonemappers
+	// also useful to scale the input to the range that the tonemapper is designed for (some require very high input values)
+	// has no effect on the curve's general shape or visual properties
+	const float exposure_bias = 2.0f;
+	const float A = 0.22f * exposure_bias * exposure_bias; // bias baked into constants for performance
+	const float B = 0.30f * exposure_bias;
+	const float C = 0.10f;
+	const float D = 0.20f;
+	const float E = 0.01f;
+	const float F = 0.30f;
+
+	vec3 color_tonemapped = ((color * (A * color + C * B) + D * E) / (color * (A * color + B) + D * F)) - E / F;
+	float white_tonemapped = ((white * (A * white + C * B) + D * E) / (white * (A * white + B) + D * F)) - E / F;
+
+	return color_tonemapped / white_tonemapped;
+}
+
+vec3 tonemap_aces(vec3 color, float white) {
+	const float exposure_bias = 0.85f;
+	const float A = 2.51f * exposure_bias * exposure_bias;
+	const float B = 0.03f * exposure_bias;
+	const float C = 2.43f * exposure_bias * exposure_bias;
+	const float D = 0.59f * exposure_bias;
+	const float E = 0.14f;
+
+	vec3 color_tonemapped = (color * (A * color + B)) / (color * (C * color + D) + E);
+	float white_tonemapped = (white * (A * white + B)) / (white * (C * white + D) + E);
+
+	return color_tonemapped / white_tonemapped;
+}
+
+vec3 tonemap_reinhard(vec3 color, float white) {
+	return (white * color + color) / (color * white + white);
+}
+
+vec3 linear_to_srgb(vec3 color) {
+	//if going to srgb, clamp from 0 to 1.
+	color = clamp(color, vec3(0.0), vec3(1.0));
+	const vec3 a = vec3(0.055f);
+	return mix((vec3(1.0f) + a) * pow(color.rgb, vec3(1.0f / 2.4f)) - a, 12.92f * color.rgb, lessThan(color.rgb, vec3(0.0031308f)));
+}
+
+#define TONEMAPPER_LINEAR 0
+#define TONEMAPPER_REINHARD 1
+#define TONEMAPPER_FILMIC 2
+#define TONEMAPPER_ACES 3
+
+vec3 apply_tonemapping(vec3 color, float white) { // inputs are LINEAR, always outputs clamped [0;1] color
+
+	if (params.tonemapper == TONEMAPPER_LINEAR) {
+		return color;
+	} else if (params.tonemapper == TONEMAPPER_REINHARD) {
+		return tonemap_reinhard(color, white);
+	} else if (params.tonemapper == TONEMAPPER_FILMIC) {
+		return tonemap_filmic(color, white);
+	} else { //aces
+		return tonemap_aces(color, white);
+	}
+}
+
+vec3 gather_glow(sampler2D tex, vec2 uv) { // sample all selected glow levels
+	vec3 glow = vec3(0.0f);
+
+	if (bool(params.glow_level_flags & (1 << 0))) {
+		glow += GLOW_TEXTURE_SAMPLE(tex, uv, 0).rgb;
+	}
+
+	if (bool(params.glow_level_flags & (1 << 1))) {
+		glow += GLOW_TEXTURE_SAMPLE(tex, uv, 1).rgb;
+	}
+
+	if (bool(params.glow_level_flags & (1 << 2))) {
+		glow += GLOW_TEXTURE_SAMPLE(tex, uv, 2).rgb;
+	}
+
+	if (bool(params.glow_level_flags & (1 << 3))) {
+		glow += GLOW_TEXTURE_SAMPLE(tex, uv, 3).rgb;
+	}
+
+	if (bool(params.glow_level_flags & (1 << 4))) {
+		glow += GLOW_TEXTURE_SAMPLE(tex, uv, 4).rgb;
+	}
+
+	if (bool(params.glow_level_flags & (1 << 5))) {
+		glow += GLOW_TEXTURE_SAMPLE(tex, uv, 5).rgb;
+	}
+
+	if (bool(params.glow_level_flags & (1 << 6))) {
+		glow += GLOW_TEXTURE_SAMPLE(tex, uv, 6).rgb;
+	}
+
+	return glow;
+}
+
+#define GLOW_MODE_ADD 0
+#define GLOW_MODE_SCREEN 1
+#define GLOW_MODE_SOFTLIGHT 2
+#define GLOW_MODE_REPLACE 3
+#define GLOW_MODE_MIX 4
+
+vec3 apply_glow(vec3 color, vec3 glow) { // apply glow using the selected blending mode
+	if (params.glow_mode == GLOW_MODE_ADD) {
+		return color + glow;
+	} else if (params.glow_mode == GLOW_MODE_SCREEN) {
+		//need color clamping
+		return max((color + glow) - (color * glow), vec3(0.0));
+	} else if (params.glow_mode == GLOW_MODE_SOFTLIGHT) {
+		//need color clamping
+		glow = glow * vec3(0.5f) + vec3(0.5f);
+
+		color.r = (glow.r <= 0.5f) ? (color.r - (1.0f - 2.0f * glow.r) * color.r * (1.0f - color.r)) : (((glow.r > 0.5f) && (color.r <= 0.25f)) ? (color.r + (2.0f * glow.r - 1.0f) * (4.0f * color.r * (4.0f * color.r + 1.0f) * (color.r - 1.0f) + 7.0f * color.r)) : (color.r + (2.0f * glow.r - 1.0f) * (sqrt(color.r) - color.r)));
+		color.g = (glow.g <= 0.5f) ? (color.g - (1.0f - 2.0f * glow.g) * color.g * (1.0f - color.g)) : (((glow.g > 0.5f) && (color.g <= 0.25f)) ? (color.g + (2.0f * glow.g - 1.0f) * (4.0f * color.g * (4.0f * color.g + 1.0f) * (color.g - 1.0f) + 7.0f * color.g)) : (color.g + (2.0f * glow.g - 1.0f) * (sqrt(color.g) - color.g)));
+		color.b = (glow.b <= 0.5f) ? (color.b - (1.0f - 2.0f * glow.b) * color.b * (1.0f - color.b)) : (((glow.b > 0.5f) && (color.b <= 0.25f)) ? (color.b + (2.0f * glow.b - 1.0f) * (4.0f * color.b * (4.0f * color.b + 1.0f) * (color.b - 1.0f) + 7.0f * color.b)) : (color.b + (2.0f * glow.b - 1.0f) * (sqrt(color.b) - color.b)));
+		return color;
+	} else { //replace
+		return glow;
+	}
+}
+
+vec3 apply_bcs(vec3 color, vec3 bcs) {
+	color = mix(vec3(0.0f), color, bcs.x);
+	color = mix(vec3(0.5f), color, bcs.y);
+	color = mix(vec3(dot(vec3(1.0f), color) * 0.33333f), color, bcs.z);
+
+	return color;
+}
+
+vec3 apply_color_correction(vec3 color, sampler3D correction_tex) {
+	return texture(correction_tex, color).rgb;
+}
+
+void main() {
+	vec3 color = textureLod(source_color, uv_interp, 0.0f).rgb;
+
+	// Exposure
+
+	if (params.use_auto_exposure) {
+		color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / params.auto_exposure_grey;
+	}
+
+	color *= params.exposure;
+
+	// Early Tonemap & SRGB Conversion
+
+	if (params.use_glow && params.glow_mode == GLOW_MODE_MIX) {
+
+		vec3 glow = gather_glow(source_glow, uv_interp);
+		color.rgb = mix(color.rgb, glow, params.glow_intensity);
+	}
+
+	color = apply_tonemapping(color, params.white);
+
+	color = linear_to_srgb(color); // regular linear -> SRGB conversion
+
+	// Glow
+
+	if (params.use_glow && params.glow_mode != GLOW_MODE_MIX) {
+
+		vec3 glow = gather_glow(source_glow, uv_interp) * params.glow_intensity;
+
+		// high dynamic range -> SRGB
+		glow = apply_tonemapping(glow, params.white);
+		glow = linear_to_srgb(glow);
+
+		color = apply_glow(color, glow);
+	}
+
+	// Additional effects
+
+	if (params.use_bcs) {
+		color = apply_bcs(color, params.bcs);
+	}
+
+	if (params.use_color_correction) {
+		color = apply_color_correction(color, color_correction);
+	}
+
+	frag_color = vec4(color, 1.0f);
+}